Identification of potential serum biomarkers of inflammation and lipid modulation that are altered by fish oil supplementation in healthy volunteers

RESEARCH ARTICLE

Identification of potential serum biomarkers of

inflammation and lipid modulation that are altered by

fish oil supplementation in healthy volunteers

Baukje de Roos1, Anouk Geelen2, 3, Karen Ross1, Garry Rucklidge1, Martin Reid1,Gary Duncan1, Muriel Caslake4, Graham Horgan5 and Ingeborg A. Brouwer2, 3

1 Department of Vascular Health, Rowett Research Institute, Aberdeen, UK2 Wageningen Centre for Food Sciences, Wageningen, The Netherlands3 Division of Human Nutrition, Wageningen University, Wageningen, The Netherlands4 Department of Vascular Biochemistry, Glasgow Royal Infirmary, Glasgow, UK5 Biomathematics and Statistics Scotland, Rowett Research Institute, Aberdeen, UK

Long chain n-3 polyunsaturated fatty acids (n-3 LCPUFA) lower risk of coronary heart disease(CHD), but mechanisms are not well understood. We used proteomics to identify human serumproteins that are altered by n-3 LCPUFA. Such proteins could identify pathways whereby theyaffect CHD. Eighty-one healthy volunteers entered a double blind randomised trial to receive3.5 g of fish oil or 3.5 g of high oleic sunflower oil daily. Serum was collected before and after6 wk of intervention. Serum was analysed by proteomics using 2-DE. Proteins that were differ-entially regulated were identified by MS. We also analysed serum apolipoprotein A1 (apo A1),high-density lipoprotein (HDL) particle size and haptoglobin. Serum levels of apo A1, apo L1,zinc-a-2-glycoprotein, haptoglobin precursor, a-1-antitrypsin precursor, antithrombin III-likeprotein, serum amyloid P component and haemopexin were significantly downregulated (allp,0.05) by fish oil compared with high oleic sunflower oil supplementation. Fish oil supple-mentation caused a significant shift towards the larger, more cholesterol-rich HDL2 particle. Thealterations in serum proteins and HDL size imply that fish oil activates anti-inflammatory andlipid modulating mechanisms believed to impede the early onset of CHD. These proteins arepotential diagnostic biomarkers to assess the mechanisms whereby fish oil protects against CHDin humans.

Received: May 14, 2007Revised: January 17, 2008

Accepted: January 18, 2008

Keywords:

Biomarkers / Fish oil / Inflammation / Lipid metabolism / Serum proteomics

Proteomics 2008, 8, 1965–1974 1965

1 Introduction

The long chain n-3 polyunsaturated fatty acids (n-3LCPUFA) eicosapentaenoic acid (EPA) and docosahexaenoicacid (DHA) are found in relatively high concentrations inoily fish and the oils derived from them. Numerous epide-miological studies suggest that consumption of fish or n-3LCPUFA protects against coronary heart disease (CHD) [1–3]. In addition, some secondary intervention studies haveshown significant benefits when n-3 LCPUFA were ad-ministered to patients who had already suffered from myo-

Correspondence: Dr. Baukje de Roos, Rowett Research Institute,Division of Vascular Health, Greenburn Road, Bucksburn, Aber-deen AB21 9SB, UKE-mail: [email protected]: 144-1224-716629

Abbreviations: apo A1, apolipoprotein; CHD, coronary heart dis-ease, DHA, docosahexaenoic acid; EPA, eicosapentaenoic acid;HDL, high density lipoprotein; LCPUFA, long-chain polyunsatu-rated fatty acids; NF-kB, nuclear factor-kB; SAP, serum amyloidP; TNF-Æ, tumour necrosis factor alpha; VLDL, very low densitylipoprotein; ZAG, zinc-a-2-glycoprotein

DOI 10.1002/pmic.200700457

© 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

1966 B. d. Roos et al. Proteomics 2008, 8, 1965–1974

cardial infarction [4]. A recent meta-analysis of observationalstudies on fish intake and CHD indicated that fish intake isassociated with a significantly lower risk of fatal and totalCHD [3]. n-3 LCPUFA can modulate numerous physiologi-cal and cellular functions involved in atherogenesis that may,either individually or collectively, underlie their beneficialeffects [5]. An important beneficial effect of n-3 LCPUFA isthe lowering of fasting plasma triglyceride concentrationsand the decrease of the postprandial response [2]. Dietaryfish oil also decreases atherosclerosis in animal models [6],which might be due to the triglyceride lowering, decreasedgrowth factor production [7], decreased inflammation [8] or acombination of these effects.

Despite the large number of studies on the potentialbeneficial effects of fish oil, many issues relating to theintake of dietary fatty acids remain unsettled, including theelucidation of their mechanism of action. This studyattempted to identify serum proteins that are either up- ordownregulated by fish oil, compared with high oleic sun-flower oil, in healthy volunteers. The discovery of such bio-marker proteins will not only enhance the understanding ofmechanisms whereby n-3 LCPUFA may affect the develop-ment of CHD, but also enable us to detect the effects of thesedietary fatty acids on the early onset of this disease inhumans in future studies.

2 Materials and methods

2.1 Human intervention study design

The design of this placebo-controlled and double-blind studyhas been described before [9]. Briefly, 81 apparently healthysubjects aged 50–70 years were stratified by habitual fishconsumption, diastolic blood pressure and sex and thenrandomised to receive either a daily dose of 3.5 g fish oil(n = 42 subjects) or high oleic sunflower oil (n = 39 subjects)(Loders Croklaan, Wormerveer, The Netherlands) during a12 wk intervention period. The oils were administered inseven soft gelatin capsules daily each containing 500 mg oiland ,0.15 mg a-tocopherol, 0.75 mg a-tocopherol and0.60 mg a-tocopherol as antioxidants. The daily dose of fishoil provided ,700 mg EPA, 560 mg DHA and 260 mg ofother n-3 LCPUFA. The placebo capsules contained mainlyoleic acid (18:1(n-9)). Compliance was confirmed by anincrease in the proportion of EPA in serum cholesteryl estersin the fish oil group compared with no change in the placebogroup [9].

2.2 Sample selection

Individual nonfasting serum samples were obtained 4 wkbefore the start of the intervention study (t = 0) and after6 wk of intervention (t = 6). Samples were kept frozen for 39–42 months at 2807C until the sample preparation for prote-omics. Sera were pooled by an external person unfamiliar

with the study design. The fish oil intervention group wasrepresented by eight pools (six pools of five samples each andtwo pools of six samples each) at weeks 0 and 6. The controlgroup was also represented by eight pools (seven pools of fivesamples each and one pool of four samples) at weeks 0 and 6.Pools were representing the same volunteers at t = 0 andt = 6.

2.3 Sample preparation for proteomics

The 32-pooled serum samples were defrosted on ice. In orderto increase the detection of low abundance proteins, weselectively removed six high abundance proteins (HSA, IgG,fibrinogen, transferrin, IgA and IgM) from each serum poolusing a Seppro™ multiaffinity protein separation system(Genway Biotech, San Diego, USA) according to the manu-facturer’s instructions. The protein concentration in theflow-through sample (depleted of the six abundant serumproteins) was determined using the Bradford assay. Protein(100 mg) per pooled sample was loaded per gel.

2.4 2-DE

One 2-DE gel per pool was run basically as described in ref.[10]. In this study we used BioRad IPG strips (pI 4–7) for theseparation of the proteins in the first dimension. Coommas-sie blue -stained gels were analysed using the automaticmatching tool within PDQuest software (BioRad), yieldingon average around 350 valid protein spots per gel. All indi-vidual matched spots were validated manually and data werenormalised within the software according to total density ingel image. Spot density values were then exported from thePDQuest software into Genstat software to perform the sta-tistical analysis as described below. A total of ten spotsrepresenting proteins of which the response from baseline inthe fish oil group significantly differed from the responsefrom baseline in the high oleic sunflower oil control groupwere excised in duplicate from the SDS-PAGE gels using therobotic BioRad spot cutter for identification. These proteinswere trypsinised using a protocol of the MassPrep Station(Micromass) and analysed by MALDI-TOF and LC-MS/MSmethods as described in ref. [10].

2.5 MALDI-TOF MS

MALDI-TOF MS was performed using an Applied Biosys-tems Voyager-DE PRO in reflectron mode. Each spectrumwas obtained using 500 shots at the appropriate laser powerand, where appropriate, spectra were accumulated and filed.A macro was applied which allowed baseline correction andde-isotoping of the peptide mass peaks. A peptide mass list ofthe most intense peaks was generated automatically and thislist was pasted into Matrix Science MASCOT, using theMSDB database during the search. We set the followingsearch criteria: allowance of zero or one missed cleavages;peptide mass tolerance at60.4 Da; trypsin as digestion en-

© 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

Proteomics 2008, 8, 1965–1974 Systems Biology 1967

zyme; carbamidomethyl modification of cysteine; methioneoxidation as partial modification and charged state asMH1[9].

2.6 LC-MS/MS analysis

The identity of all spots were also analysed by an ‘Ultimate’nanoLC system (LC Packings, Camberly, Surrey, UK) using aC18 PepMap 100 nanocolumn, 15 cm675 mm id, 3 mm,100 Å (LC Packings). The nanoLC system was operated witha column flow rate set at 0.3 mL/min, a ‘Famos’ autosamplerset to an injection volume of 5 mL and a ‘Switchos’ micro-column switching device set at a flow rate 0.03 mL/min. Thesolvent used by the ‘Switchos’ was 0.1% formic acid. We used2% ACN and 0.1% formic acid (A) and 80% ACN and 0.08%formic acid (B) as HPLC grade solvents, with the gradientstarting at 5% B, going to 50% B over 30 min, then rampingto 80% B over a further 2 min and then held for 10 min. Thesystem was equilibrated at 95% A for 9 min prior to injectionof subsequent samples.

The MS was performed using a Q-Trap (Applied Biosys-tems/MDS Sciex, Warrington, UK) triple quadrupole massspectrometer fitted with a nanospray ion source, where Q3was operated as a linear IT (LIT). The nanospray needlevoltage was set at 1850 V. OFN (oxygen free nitrogen) wasused as the curtain gas and the collision gas. In the surveyscan mode, the mass range in Q1 was set to m/z 400–1200with a scan rate of 4000 amu/s. The criteria for selection ofions for the fragmentation (Q2) were ions of 105 cps (countsper second) or above. The collision energy was compounddependent (set to a maximum of 80 eV). The trap fill time(Q3) was 250 ms and the scan rate was 1000 amu/s. The TICdata were submitted for database searching using the MAS-COT search engine (Matrix Science) using the MSDB data-base with the following search criteria: allowance of zero orone missed cleavages; peptide mass tolerance of 61 Da;fragment mass tolerance of 60.8 Da, trypsin as digestionenzyme; carabamidomethyl modification of cysteine;methionine oxidation as partial modification; and chargedstate as MH1.

2.7 Serum analysis of apolipoprotein A1 (apo A1) and

high density lipoprotein (HDL) subfractions

Apo A1 was analysed by immunoturbidimetry using a com-mercial kit (Wako, Alpha Laboratories, UK). Three additionalspots, believed to be positional variants of apo A1 protein asreported previously [11], were excised from the gels, trypsi-nised and identified using MALDI-TOF and LC-MS/MS asdescribed above and changes in spot densities were analysedas above. HDL (d = 1.063–1.21 g/mL) was isolated fromserum by sequential ultracentrifugation. Electrophoresis wascarried out on 2–30% polyacrylamide gels (Alamo gels, SanAntonia, TX) to determine the particle size. Each gel wasstandardised using high molecular weight markers (Amers-

ham Biosciences, UK) and scanned by laser densitometry(BioRad Multianalist™ version 1.1, Hercules, Canada) [12].

2.8 Serum analysis of haptoglobin

Haptoglobin levels in serum were analysed by ELISA (ICL,Newberg, OR, USA) according to the manufacturer’sinstructions.

2.9 Statistics

The unpaired Student’s t-test was used to determine differ-ences in responses between the fish oil and the controltreatment during the 6-wk intervention period using Genstatsoftware. The analysis of multiple hypotheses testing wasdone by determining the q-values [13] within Genstat. q-Values estimate the probability that a correlation that iscalled significant, is false positive.

3 Results

Proteomics revealed that the levels of the following tenserum proteins were significantly (all p,0.05, q,0.76)downregulated by 6 wk of fish oil supplementation as com-pared to 6 wk of high oleic sunflower oil supplementation:apo A1, apo L1, zinc-a-2-glycoprotein (ZAG), three post-translationally modified forms of haptoglobin precursor, a-1-antitrypsin precursor, antithrombin III-chain L, serum amy-loid P (SAP) component and haemopexin (Tables 1 and 2,Fig. 1).

In addition to the one form of apo A1 protein of whichlevels were significantly decreased upon fish oil supple-mentation as compared with high oleic sunflower supple-mentation, levels of three other positional variants of apo A1protein, that have been identified previously [11], were notaffected by the dietary intervention (Tables 1 and 2, Fig. 1).

Validation studies in the pooled serum samples revealedthat fish oil supplementation lowered serum levels of apo A1(mean 6 SD: 24.34 6 7.95 mg/dL) as compared with higholeic sunflower supplementation (210.50 6 8.16 mg/dL)(p = 0.15), although this decrease did not reach statisticalsignificance. However, fish oil supplementation did cause asignificant shift towards the larger, more cholesterol-richHDL2 particle (Fig. 2). Fish oil supplementation did notlower serum haptoglobin levels upon fish oil supplementa-tion (mean 6 SD: 0.25 6 2.21 mg/mL), as compared withhigh oleic sunflower oil supplementation (mean 6 SD:20.41 6 3.91) (p = 0.34).

4 Discussion

In this study we identified ten proteins that were signifi-cantly up- or downregulated in healthy subjects consumingfish oil supplements for 6 wk compared with a group con-

© 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

1968 B. d. Roos et al. Proteomics 2008, 8, 1965–1974

Figure 1. Representative 2-DEgels of the two interventiongroups after 6 wk of interventionindicating the differential proteinlevels of apo A1, apo L1, ZAG,three positional variants of hap-toglobin precursor, a-1-anti-trypsin precursor, antithrombinIII precursor, SAP component,hemopexin and gelsolin, asidentified by MALDI-TOF MS andLC-MS/MS and described in Sec-tion 2. Black arrows indicateserum proteins of which levelswere significantly altered by fishoil supplementation. Whitearrows indicate the differentpositional variants of apo A1 thatwere not significantly altered byfish oil supplementation.

suming high oleic sunflower oil supplements. The functionsof the proteins could be categorised as modulators ofinflammation, as modulators of lipid/lipoprotein metabo-lism or both.

4.1 Modulators of inflammation

4.1.1 Haptoglobin, haemopexin and Æ-1-antitrypsin

Fish oil significantly lowered three post-translationally mod-ified forms of haptoglobin precursor, haemopexin and a-1-antitrypsin protein levels in serum. However, serum levels ofthe mature haptoglobin molecule (the tetramer consisting oftwo alpha and two beta chains that arise from proteolyticprocessing of the same precursor) were not altered. Hap-toglobin, haemopexin and a-1-antitrypsin are positive acute-phase proteins of which serum concentrations normallyincrease during an acute phase response. The acute-phaseresponse occurs following a wide variety of insults to thebody including trauma, infection, inflammation and cancer.In humans, acute-phase proteins are produced in the liver. Inin vitro experiments its production is largely regulated by thepotent proinflammatory cytokines interleukin-6 (IL-6) andtumour necrosis factor alpha (TNF-a). Thus an increase inthe acute-phase protein response in vivo may be taken as in-direct evidence of proinflammatory cytokine production [14].The downregulation of three positive acute-phase proteinsindicates that consumption of fish oil may be associated witha decrease in the acute-phase protein response, possibly be-cause of a modulation of the inflammatory response. Long-chain n-3 LCPUFA decrease inflammation by lowering theproduction of chemo-attractants, growth factors, adhesionmolecules, inflammatory eicosanoids and inflammatorycytokines in some studies [15] but not in others [5]. Resultsfrom in vivo studies assessing the effects of fish oil supple-

mentation on circulating levels of soluble adhesion mole-cules in humans have not been conclusive [5]. Overall, invitro studies show that n-3 fatty acids, particularly DHA, de-crease expression of vascular cell adhesion molecule-1(VCAM-1) on the vascular endothelium and reduced leuko-cyte rolling and adhesion to the endothelium. The produc-tion of proinflammatory cytokines and the acute phase re-sponse is controlled to some extent by the transcription fac-tor nuclear factor-kB (NF-kB) [16]. One study [17] showedthat DHA inhibits the activation of the NF-kB system oftranscription factors. The mechanism by which n-3 LCPUFAdiminish the activation of the NF-kB system in response tocytokines is still unknown.

4.2 Modulators of lipid metabolism

4.2.1 Apolipoprotein A1

Proteomics analysis revealed that fish oil significantlydecreased levels of one positional variant of serum apo A1,the major protein component of high-density lipoproteins(HDL) (Table 1, Fig. 1). The fact that fish oil supplementationonly significantly decreased protein levels of one of the fourpositional variants of apo A1, whereas levels of three otherpositional variants of apo A1 were unaffected, may explainwhy fish oil supplementation did not change overall serumlevels of apo A1 as measured by immunoturbidimetry (Table2, Fig. 1).

Any decrease in apo A1 protein may indicate either a de-crease in the number of HDL particles or a change in sizeand/or composition of HDL particles. A previous humanintervention study showed no effect of fish oil treatmentcompared with olive oil treatment on overall plasma apo A1levels. However, within-group decreases in apo A1/HDL-cholesterol in the EPA and DHA groups were observed, with

© 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

Proteomics 2008, 8, 1965–1974 Systems Biology 1969

Table 1. Serum proteins that were significantly downregulated by 6 wk of fish oil supplementation compared to 6 wk of high oleic sun-flower oil supplementation, as assessed by proteomicsa)

SSP Protein Accession no. Fish oil High oleic sunflower oil Fish oil versus higholeic sunflower oil

p

Before After Before After Difference (95% CI)

3204 Apo A1 LPHUA1 1348 6 609 984 6 260 901 6 328 1100 6 272 2546 (21070; 258) 0.032205 Apo A1 LPHUA1 7421 6 3180 6483 6 1367 5687 6 2391 5717 6 1228 2969 (23620; 1682) 0.443201 Apo A1 LPHUA1 36 864 6 11 156 34 656 6 5241 28 328 6 11 701 32 318 6 8375 26197 (217 780; 5386) 0.274201 Apo A1 LPHUA1 1662 6 732 1274 6 224 1216 6 411 1443 6 258 2615 (21294; 63) 0.074413 Apo L1 AAK20210 55 6 77 19 6 42 30 6 53 50 6 67 257 (2113 ; 0) 0.051401 ZAG 1ZAGD 683 6 795 409 6 491 415 6 580 1140 6 861 21000 (21946; 254) 0.044403 Haptoglobin

precursorAAH70299 1877 6 633 1363 6 349 1275 6 460 1536 6 436 2774 (21404; 2145) 0.02

4404 Haptoglobinprecursor

Q6NSB4_HUMAN 938 6 591 630 6 144 626 6 201 825 6 168 2507 (2983; 230) 0.04

5201 Haptoglobinprecursor

HPHU2 8265 6 3558 6793 6 2479 5994 6 2234 7896 6 3452 23374 (26734; 214) 0.05

1502 a-1-Antitrypsinprecursor

ITHU 12 295 6 6739 8964 6 4278 10 702 6 3807 12 719 6 5369 25348 (210 124; 2572) 0.03

2603 AntithrombinIII–chain L

1BR8L 494 6 284 294 6 225 219 6 139 210 6 117 2192 (2371; 212) 0.04

4302 SAP component YLHUP 1117 6 355 895 6 278 883 6 215 944 6 119 2296 (2586; 26) 0.055805 Haemopexin HSHEPEXR 5387 6 2393 3722 6 1042 3461 6 1405 4624 6 1359 22827 (25255; 2400) 0.03

a) Values represent mean spot densities, as determined by the PDQuest software, 6 SD. We calculated the response to fish oil andhigh oleic sunflower oil by subtracting pretreatment values from treatment values. Differences in responses between fish oil andhigh oleic acid sunflower oil were tested using the unpaired Student’s t-test. SSP = spot identity number within PDQuestmatchset.

Figure 2. Mean percentage change in HDL particle size uponsupplementation with fish oil (black bars) and upon interventionwith high oleic sunflower oil (white bars),6SD. Differences inmean percentage change were tested using the one-tailedunpaired Student’s t-test, * p,0.05.

the 9% decrease in the DHA group reaching significance[18]. This particular result is suggestive of an effect of sup-plementation on HDL particle size, with a shift towards thelarger, more cholesterol-rich HDL2 particle. Indeed, weshowed that fish oil supplementation significantly decreasedthe amount of the smaller HDL3a and HDL3b particles andsignificantly increased the amounts of the larger HDL2aparticles. Increased HDL particle size is thought to serve as amarker for more efficient reverse cholesterol transport and

lower CHD risk in humans [19]. Indeed, Mori et al. [20] andRambjør et al. [21] observed significant increases in HDL2

following DHA but not EPA supplementation.

4.2.2 Apolipoprotein L1

This is the first time that fish oil consumption has beenfound to lower serum levels of apo L1. Apo L1, a 42 kDaprotein, is found in the serum of humans and African greenmonkeys (but not of other animals like dogs, rabbits, pigs,rats and mice), mainly associated with large-diameter HDLparticles. The apo L1 gene has a putative sterol response ele-ment in its promoter, and sequence analysis has revealedconserved amphipatic helices, suggesting that this proteinmight be involved in lipid metabolism [22].

Apo L1 protein levels were positively associated withplasma triglycerides in both normolipidaemic and dyslipi-daemic human subjects [23, 24], which indicates that the de-crease in serum apo L1 levels might play a role in the hypo-triglyceridaemic effects of fish oil. However, because a smallfraction of total apo L1 has been found on the very low-den-sity lipoprotein (VLDL) particle, it is possible that the associ-ation of apo L1 with plasma triglycerides is directly related toVLDL particle numbers or composition. Furthermore, theexpression of the apo L1 gene has been found to be upregu-lated by TNF-a in endothelial cells [22], suggesting that the

© 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

1970 B. d. Roos et al. Proteomics 2008, 8, 1965–1974

Table 2. Measures of confidence for protein identification and characterisation by MALDI-TOF and MS/MS analysis

SSP Protein Theor. Mr

(kDa)Exper. Mr

(kDa)Proteinscorea)

MS/MS PMF

No.matchedpeptides

Peptide sequences Peptidecharges

No.matchedpeptides

Sequencecoverage(%)

3204 Apo A1 28.1 24.8 342 13 AKPALEDLRLSPLGEEMRATEHLSTLSEKDEPPQSPWDRDLATVYVDVLKVQPYLDDFQKWQEEMELYRWQEEMELYR 1 Ox (M)THLAPYSDELRVSFLSALEEYTKDYVSQFEGSALGKLLDNWDSVTSTFSKLLDNWDSVTSTFSK

12121212121212121212121312

15 57

2205 Apo A1b) 28.1 24.8 478 16 AKPALEDLRLSPLGEEMRATEHLSTLSEKDEPPQSPWDRDLATVYVDVLKVQPYLDDFQKWQEEMELYRWQEEMELYR 1 Ox (M)THLAPYSDELRVSFLSALEEYTKDYVSQFEGSALGKVEPLRAELQEGARLLDNWDSVTSTFSKLLDNWDSVTSTFSKDSGRDYVSQFEGSALGKEQLGPVTQEFWDNLEK

12121212121212121212121312131312

14 54

3201 Apo A1b) 28.1 24.8 638 19 AKPALEDLRLSPLGEEMRLEALKENGGARATEHLSTLSEKDEPPQSPWDRVQPYLDDFQKWQEEMELYRTHLAPYSDELRETEGLRQEMSKQEMSKDLEEVK 1 Ox (M)VSFLSALEEYTKDYVSQFEGSALGKKWQEEMELYRAKVQPYLDDFQKVKDLATVYVDVLKLLDNWDSVTSTFSKLLDNWDSVTSTFSKEQLGPVTQEFWDNLEKLREQLGPVTQEFWDNLEK

12121212121212121213121212121212131213

12 46

4201 Apo A1b) 28.1 24.8 327 11 AKPALEDLRATEHLSTLSEKDLATVYVDVLKVQPYLDDFQK

12121212

12 49

© 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

Proteomics 2008, 8, 1965–1974 Systems Biology 1971

Table 2. Continued

SSP Protein Theor. Mr

(kDa)Exper. Mr

(kDa)Proteinscorea)

MS/MS PMF

No.matchedpeptides

Peptide sequences Peptidecharges

No.matchedpeptides

Sequencecoverage(%)

VQPYLDDFQKWQEEMELYRWQEEMELYR 1 Ox (M)THLAPYSDELRVQPYLDDFQKKDYVSQFEGSALGKKWQEEMELYR 1 Ox (M)LLDNWDSVTSTFSK

1212121213121312

4413 Apo L1 43.8 40.8 168 4 SELEDNIRLNILNNNYKVNEPSILEMSRVNEPSILEMSR 1 Ox (M)

12121212

No identification

1401 ZAG 31.9 40.5 136 3 AGEVQEPELRAYLEEECPATLRYYYDGKDYIEFNK

121213

17 35

4403 Haptoglobinprecursor

38.7 39.1 81 10 GSFPWQAKILGGHLDAKVGYVSGWGRVMPICLPSK 1 Ox (M)VTSIQDWVQKDIAPTLTLYVGKSCAVAEYGVYVKYVMLPVADQDQCIRYVMLPVADQDQCIR 1 Ox (M)YVMLPVADQDQCIR 1 Ox (M)

12121212121212131312

9 34

4404 Haptoglobinprecursor

31.6 36.6 68 10 ILGGHLDAKVGYVSGWGRVMPICLPSK 1 Oxidation (M)VTSIQDWVQKDIAPTLTLYVGKSCAVAEYGVYVKYVMLPVADQDQCIR 1 Ox (M)YVMLPVADQDQCIR 1 Ox (M)VMPICLPSKDYAEVGR 1 Ox (M)SPVGVQPILNEHTFCAGMSK 1 Ox (M)

12121212121213121313

14 34

5201 Haptoglobinprecursor

45.8 19.6 387 10 TEGDGVYTLNDKLPECEAVCGKPKLPECEAVCGKPKTEGDGVYTLNDKKTEGDGVYTLNDKKLRTEGDGVYTLNDKLRTEGDGVYTLNDKLRTEGDGVYTLNNEKLRTEGDGVYTLNNEKAVGDKLPECEAVCGKPK

12121312131213121313

No identification

1502 a-1-Antitrypsinprecursor

46.9 52.0 72 13 FLENEDRQINDYVEKSVLGQLGITKLSITGTYDLKLGMFNIQHCK 1 Ox (M)LQHLENELTHDIITK

121212121213

13 35

© 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

1972 B. d. Roos et al. Proteomics 2008, 8, 1965–1974

Table 2. Continued

SSP Protein Theor. Mr

(kDa)Exper. Mr

(kDa)Proteinscorea)

MS/MS PMF

No.matchedpeptides

Peptide sequences Peptidecharges

No.matchedpeptides

Sequencecoverage(%)

VFSNGADLSGVTEEAPLKDTEEEDFHVDQVTTVKDTEEEDFHVDQVTTVKLYHSEAFTVNFGDTEEAKLYHSEAFTVNFGDTEEAKKGTEAAGAMFLEAIPMSIPPEVK 1 Ox (M)TLNQPDSQLQLTTGNGLFLSEGLK

12121313131313

2603 AntithrombinIII–chain L

53.0 57.6 75/579c) 11 IEDGFSLKRVWELSKFRIEDGFSLKDDLYVSDAFHKTSDQIHFFFAKEVPLNTIIFMGREVPLNTIIFMGR 1 Ox (M)VAEGTQVLELPFKDIPMNPMCIYR 1 2 Ox (M)ADGESCSASMMYQEGKATEDEGSEQKIPEATNR

1212121212121212121213

14 25

4302 SAP component 25.5 27.8 306 5 DNELLVYKQGYFVEAQPKGYVIIKPLVWVIVLGQEQDSYGGKIVLGQEQDSYGGKFDR

1212121213

No identification

5805 Hemopexin 49.9 70.1 343 7 VWVYPPEKVDGALCMEKVDGALCMEK 1 Ox (M)GGYTLVSGYPKDYFMPCPGR 1 Ox (M)NFPSPVDAAFRYYCFQGNQFLR

12121212121212

9 28

a) Probability-based Mowse score for MS/MS-based identifications: individual ions scores .35 indicate identity or extensive homology(p,0.05). Protein score is 210*log (P), where P is the probability that the observed match is a random event.

b) Protein levels were not significantly different upon fish oil supplementation as compared to high oleic sunflower oil supplementation.c) 75 is the protein score for PMF-based identification; 579 is the protein score for MS/MS-based identification.

decreased serum levels of apo L1 found in our study may alsocontribute to the anti-inflammatory action of fish oil.

4.2.3 Zinc-Æ-2-glycoprotein

Adipocytes express and secrete a wide range of proteinstermed adipokines, such as leptin, adiponectin, resistin, apoE, TNF-a and IL-6. These adipocyte-derived proteins acteither in an autocrine/paracrine manner to locally regulateadipocyte metabolism or as endocrine signals at distant sitesin relation to energy homeostasis and other physiologicalprocesses, such as insulin resistance and the inflammatoryresponse. ZAG is considered a novel adipokine; it is a solubleprotein that is present in serum and other body fluids. The

enzyme is expressed and secreted by human adipocytes,causing lipid degradation in adipocytes through stimulationof adipocyte adenylate cyclase, thereby activating the enzymehormone sensitive lipase [25].

Previous research has shown that EPA is effective in sta-bilising body weight in patients with advanced pancreaticcancer [26]. EPA may preserve adipose tissue mass incachexia by attenuating the upregulation of ZAG expressionin white and brown adipose tissue [27]. Likewise, the de-crease in serum ZAG levels after intervention with fish oil(Table 1) could indicate a potential alternative mechanism forthe lowering in plasma/serum triglycerides through a de-crease in the activity of hormone sensitive lipase, and a sub-sequent decrease in levels of plasma/serum free fatty acids.

© 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

Proteomics 2008, 8, 1965–1974 Systems Biology 1973

4.2.4 Serum amyloid P-component precursor

SAP component protein decreased in serum upon con-sumption of fish oil (Table 1). SAP binds to various lipopro-tein ligands in a calcium-dependent manner and the amountof SAP in human plasma/serum positively correlates withthe level of VLDL and negatively correlates with the level ofHDL. In addition, SAP is specifically localised in humanatherosclerotic lesions, and the amount of SAP in the humanaorta correlates positively with the degree of atherosclerosis.Taken together, these findings suggest that SAP may play arole in the pathogenesis of atherosclerosis, probably relatedto the metabolism of lipoproteins [28].

4.3 Modulator of coagulation

4.3.1 Antithrombin III-chain L

Antithrombin is a serine proteinase inhibitor (serpin) thatcontrols the process of coagulation. It is the most importantanticoagulant molecule in mammalian circulation systems,controlled by its interaction with the cofactor, heparin, whichaccelerates its interaction with target proteases, such asthrombin and factor Xa. Antithrombin-III inhibits thrombinas well as factors IXa, Xa and Xia. Since the 1970s, fish oil hasbeen studied as a nutritional component with antithromboticpotential. Some trials point to a moderate reduction by fishoil of the plasma levels of fibrinogen and coagulation factorsV, VII and X, whereas other studies fail to detect this [29]. Inview of its antithrombotic potential, the finding that fish oilreduces the amount of antithrombin III-chain L protein cancurrently not be explained.

5 Conclusion

In this study we observed regulation of a number of serumproteins which suggested that fish oil may inhibit the acutephase response, and at the same time modulates lipid me-tabolism to establish a less atherogenic lipoprotein profile,i.e. lowering levels of serum triglycerides and increasing theamount of HDL2 particles, as observed in previous studies.Our proteomics approach does not allow us to concludewhether these effects are shared by one common pathway orrepresent different pathways. The q-values of the proteinsdownregulated by fish oil intervention imply that a propor-tion of the spots found significant might be explained aschance findings. The risk of a high rate of false positiveresults has to be addressed in proteomics studies [30]. How-ever, since the downregulated proteins detected in this studyare metabolically related to lipoprotein metabolism andinflammation, which are also pathways affected by con-sumption of fish oils [5], it is unlikely that such joint changesoccur by chance.

After the onset of an acute phase response, serum tri-glycerides can markedly increase [31, 32]. This effect is

established by an increased production of triglycerides andVLDL in the liver [31, 32], a decreased activity of lipoproteinlipase, a key enzyme in the hydrolysis of triglyceride-richlipoproteins [32] and a decreased hepatic fatty acid oxidation[33]. Albeit not proven, an inhibition of the acute phase re-sponse may subsequently lower serum triglycerides via thesame pathways. Alternatively, lipoprotein lipase can sup-press TNF-a-induced gene expression via inhibition of NF-kB activity in human aortic endothelial cells [34], indicatingthat an enhanced triglyceride catabolism may precede ananti-inflammatory response.

The serum proteins identified in this study revealed aconcomitant effect on inflammatory and lipid-modulatingpathways, both believed to play important roles in preventingthe early onset of CHD. Future studies could use these pro-teins as novel serum biomarkers to assess how n-3 LCPUFAaffects these pathways in order to defer the development ofCHD in humans.

Funding for the intervention study was obtained by theWageningen Centre for Food Sciences. Wageningen Centre forFood Sciences is an alliance of major Dutch food industries, TNONutrition and Food Research, Maastricht University andWageningen University and Research Centre, the Netherlandswith financial support by the Dutch Government. Funding for theproteomics analysis was obtained by the Scottish GovernmentRural and Environment Research and Analysis Directorate(RERAD).

The authors have declared no conflict of interest.

6 References

[1] Burr, M. L., Fehily, A. M., Gilbert, J. F., Rogers, S. et al., Effectsof changes in fat, fish, and fibre intakes on death and myo-cardial reinfarction: Diet and reinfarction trial (DART). Lancet1989, 2, 757–761.

[2] Harris, W. S., n-3 fatty acids and lipoproteins: Comparison ofresults from human and animal studies. Lipids 1996, 31, 243–252.

[3] Whelton, S. P., He, J., Whelton, P. K., Muntner, P., Meta-analy-sis of observational studies on fish intake and coronary heartdisease. Am. J. Cardiol. 2004, 93, 1119–1123.

[4] Gruppo Italiano per lo Studio della Sopravvivenza nell’Infartomiocardico. Dietary supplementation with n-3 poly-unsaturated fatty acids and vitamin E after myocardialinfarction: Results of the GISSI-Prevenzione trial. Lancet 1999,354, 447–455.

[5] de Roos, B., Mercer, D. K., Wainwright, C., Thies, F., Effect oflong chain n-3 PUFA on endothelial activation, endothelialfunction and atheromatous plaque stability. Curr. Nutr. FoodSci. 2005, 1, 167–177.

[6] Mortensen, A., Hansen, B. F., Hansen, J. F., Frandsen, H. et al.,Comparison of the effects of fish oil and olive oil on bloodlipids and aortic atherosclerosis in Watanabe heritablehyperlipidaemic rabbits. Br. J. Nutr. 1998, 80, 565–573.

© 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

1974 B. d. Roos et al. Proteomics 2008, 8, 1965–1974

[7] Wallace, J. M., Turley, E., Gilmore, W. S., Strain, J. J., Dietaryfish oil supplementation alters leukocyte function and cyto-kine production in healthy women. Arterioscler. Thromb.Vasc. Biol. 1995, 15, 185–189.

[8] Endres, S., Ghorbani, R., Kelley, V. E., Georgilis, K. et al., Theeffect of dietary supplementation with n-3 polyunsaturatedfatty acids on the synthesis of interleukin-1 and tumornecrosis factor by mononuclear cells. N. Engl. J Med. 1989,320, 265–271.

[9] Geelen, A., Brouwer, I. A., Zock, P. L., Kors et al., (N-3) fattyacids do not affect electrocardiographic characteristics ofhealthy men and women. J. Nutr. 2002, 132, 3051–3054.

[10] de Roos, B., Duivenvoorden, I., Rucklidge, G., Reid, M. et al.,Response of apolipoprotein E*3-Leiden transgenic mice todietary fatty acids: Combining liver proteomics with physi-ological data. FASEB J. 2005, 19, 813–815.

[11] Anderson, L., Anderson, N. G., High resolution two-dimen-sional electrophoresis of human plasma proteins. Proc.Natl. Acad. Sci. USA 1977, 74, 5421–5425.

[12] Blanche, P. J., Gong, E. L., Forte, T. M., Nichols, A. V., Char-acterization of human high-density lipoproteins by gradientgel electrophoresis. Biochim. Biophys. Acta 1981, 665, 408–419.

[13] Storey, J. D., Tibshirani, R., Statistical significance for ge-nomewide studies. Proc. Natl. Acad. Sci. USA 2003, 100,9440–9445.

[14] Gabay, C., Kushner, I., Acute-phase proteins and other sys-temic responses to inflammation. N. Engl. J. Med. 1999, 340,448–454.

[15] Calder, P. C., n-3 Fatty acids and cardiovascular disease: Evi-dence explained and mechanisms explored. Clin. Sci. (Lond)2004, 107, 1–11.

[16] Beauparlant, P., Hiscott, J., Biological and biochemical in-hibitors of the NF-kappa B/Rel proteins and cytokine syn-thesis. Cytokine Growth Factor Rev. 1996, 7, 175–190.

[17] Weber, C., Erl, W., Pietsch, A., Danesch, U., Weber, P. C.,Docosahexaenoic acid selectively attenuates induction ofvascular cell adhesion molecule-1 and subsequent mono-cytic cell adhesion to human endothelial cells stimulated bytumor necrosis factor-alpha. Arterioscler. Thromb. Vasc.Biol. 1995, 15, 622–628.

[18] Buckley, R., Shewring, B., Turner, R., Yaqoob, P., Minihane,A. M., Circulating triacylglycerol and apoE levels in responseto EPA and docosahexaenoic acid supplementation in adulthuman subjects. Br. J. Nutr. 2004, 92, 477–483.

[19] Alagona, C., Soro, A., Ylitalo, K., Salonen, R. et al., A lowhigh density lipoprotein (HDL) level is associated with car-otid artery intima-media thickness in asymptomatic mem-bers of low HDL families. Atherosclerosis 2002, 165, 309–316.

[20] Mori, T. A., Burke, V., Puddey, I. B., Watts et al., Purifiedeicosapentaenoic and docosahexaenoic acids have differ-ential effects on serum lipids and lipoproteins, LDL particlesize, glucose, and insulin in mildly hyperlipidemic men. Am.J. Clin. Nutr. 2000, 71, 1085–1094.

[21] Rambjor, G. S., Walen, A. I., Windsor, S. L., Harris, W. S.,Eicosapentaenoic acid is primarily responsible for hypo-

triglyceridemic effect of fish oil in humans. Lipids 1996, 31,S45–S49.

[22] Monajemi, H., Fontijn, R. D., Pannekoek, H., Horrevoets, A.J., The apolipoprotein L gene cluster has emerged recentlyin evolution and is expressed in human vascular tissue.Genomics 2002, 79, 539–546.

[23] Duchateau, P. N., Movsesyan, I., Yamashita, S., Sakai, N. etal., Plasma apolipoprotein L concentrations correlate withplasma triglycerides and cholesterol levels in normolipi-demic, hyperlipidemic, and diabetic subjects. J. Lipid Res.2000, 41, 1231–1236.

[24] Albert, T. S., Duchateau, P. N., Deeb, S. S., Pullinger et al.,Apolipoprotein L-I is positively associated with hyperglyce-mia and plasma triglycerides in CAD patients with low HDL.J. Lipid Res. 2005, 46, 469–474.

[25] Bao, Y., Bing, C., Hunter, L., Jenkins, J. R. et al., Zinc-alpha2-glycoprotein, a lipid mobilizing factor, is expressed andsecreted by human (SGBS) adipocytes. FEBS Lett. 2005, 579,41–47.

[26] Wigmore, S. J., Barber, M. D., Ross, J. A., Tisdale, M. J. et al.,Effect of oral eicosapentaenoic acid on weight loss inpatients with pancreatic cancer. Nutr. Cancer 2000, 36, 177–184.

[27] Russell, S. T., Tisdale, M. J., Effect of eicosapentaenoic acid(EPA) on expression of a lipid mobilizing factor in adiposetissue in cancer cachexia. Prostaglandins Leukot. Essent.Fatty Acids 2005, 72, 409–414.

[28] Li, X. A., Yutani, C., Shimokado, K., Serum amyloid P com-ponent associates with high density lipoprotein as well asvery low density lipoprotein but not with low density lipo-protein. Biochem. Biophys. Res. Commun. 1998, 244, 249–252.

[29] Vanschoonbeek, K., Feijge, M. A., Paquay, M., Rosing, J. etal., Variable hypocoagulant effect of fish oil intake inhumans: Modulation of fibrinogen level and thrombin gen-eration. Arterioscler. Thromb. Vasc. Biol. 2004, 24, 1734–1740.

[30] Biron, D. G., Brun, C., Lefevre, T., Lebarbenchon et al., Thepitfalls of proteomics experiments without the correct use ofbioinformatics tools. Proteomics 2006, 6, 5577–5596.

[31] Carpentier, Y. A., Scruel, O., Changes in the concentrationand composition of plasma lipoproteins during the acutephase response. Curr. Opin. Clin. Nutr. Metab. Care 2002, 5,153–158.

[32] Khovidhunkit, W., Kim, M. S., Memon, R. A., Shigenaga, J.K. et al., Effects of infection and inflammation on lipid andlipoprotein metabolism: Mechanisms and consequences tothe host. J. Lipid Res. 2004, 45, 1169–1196.

[33] Kim, M. S., Shigenaga, J. K., Moser, A. H., Feingold, K. R.,Grunfeld, C., Suppression of estrogen-related receptor{alpha} and medium-chain acyl-coenzyme A dehydrogenasein the acute-phase response. J. Lipid Res. 2005, 46, 2282–2288.

[34] Kota, R. S., Ramana, C. V., Tenorio, F. A., Enelow, R. I., Rutle-dge, J. C., Differential effects of lipoprotein lipase on tumornecrosis factor-alpha and interferon-gamma-mediated geneexpression in human endothelial cells. J. Biol. Chem. 2005,280, 31076–31084.

© 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com