+ All Categories
Home > Documents > Proteome Signature of IUGR From Cord Blood Serum

Proteome Signature of IUGR From Cord Blood Serum

Date post: 11-Jan-2016
Category:
Upload: romana-masnikosa
View: 219 times
Download: 0 times
Share this document with a friend
Description:
proteome from cord blood serum of IUGR fetuses
Popular Tags:
13
Electrophoresis 2012, 33, 1881–1893 1881 Manja W ¨ olter 1 Claudia R ¨ ower 1 Cornelia Koy 1 Toralf Reimer 2 Werner Rath 3 Ulrich Pecks 3 Michael O. Glocker 1 1 Proteome Center Rostock, Medical Faculty and Natural Science Faculty, University of Rostock, Rostock, Germany 2 Department of Obstetrics and Gynecology, Medical Faculty, University of Rostock, Clinic Suedstadt, Rostock, Germany 3 Department of Obstetrics and Gynecology, Medical Faculty, RWTH Aachen University, Germany Received January 1, 2012 Revised January 31, 2012 Accepted February 16, 2012 Research Article A proteome signature for intrauterine growth restriction derived from multifactorial analysis of mass spectrometry-based cord blood serum profiling Intrauterine growth restriction (IUGR) is defined as a condition in which the fetus does not reach its genetically given growth potential, resulting in low birth weight. IUGR is an important cause of perinatal morbidity and mortality, thus contributing substantially to medically indicated preterm birth in order to prevent fetal death. We subjected um- bilical cord blood serum samples either belonging to the IUGR group (n = 15) or to the control group (n = 15) to fractionation by affinity chromatography using a bead system with hydrophobic interaction capabilities. So prepared protein mixtures were analyzed by MALDI-TOF mass spectrometric profiling. The six best differentiating ion signals at m/z 8205, m/z 8766, m/z 13 945, m/z 15 129, m/z 15 308, and m/z 16 001 were collectively assigned as IUGR proteome signature. Separation confidence of our IUGR proteome sig- nature reached a sensitivity of 0.87 and a specificity of 0.93. Assignment of ion signals in the mass spectra to specific proteins was substantiated by SDS-PAGE in conjunction with peptide mass fingerprint analysis of cord blood serum proteins. One constituent of this proteome signature, apolipoprotein C-III 0 , a derivative lacking glycosylation, has been found more abundant in the IUGR cord blood serum samples, irrespective of gestational age. Hence, we suggest apolipoprotein C-III 0 as potential key-marker of the here proposed IUGR proteome signature, as it is a very low-density lipoprotein (VLDL) and high-density lipoprotein (HDL) member and as such involved in triglyceride metabolism that itself is discussed as being of importance in IUGR pathogenesis. Our results indicate that subtle alterations in protein glycosylation need to be considered for improving our understanding of the pathomechanisms in IUGR. Keywords: Affinity–mass spectrometry / IUGR / Cord blood analysis / MALDI-TOF MS / Multiparametric profiling / Proteome signature DOI 10.1002/elps.201200001 Correspondence: Dr. Michael O. Glocker, Proteome Center Ros- tock, University of Rostock, Schillingallee 69, D-18057 Rostock, Germany E-mail: [email protected] Fax: +49-381-494-4932 Abbreviations: AUC, area under curve; CHCA, alpha- cyano-4-hydroxy cinnamic acid; CTRL, control; DHB, 2,5- dihydroxybenzoic acid; HDL, high-density lipoproteins; IUGR, intrauterine growth restriction; [M+H] + , protonated molecu- lar ion; MRM, multi-reaction monitoring; m/z, mass to charge ratio; n-OGP, n-octylglucopyranoside; ROC, receiver operator characteristics; SGA, small for gestational age; TIC, total ion count; VLDL, very low-density lipoproteins 1 Introduction Intrauterine growth restriction (IUGR) affects about 3 to 8% of all pregnancies. It is defined as a condition in which the fetus does not reach its genetically given growth potential, resulting in low birth weight. The pathogenesis of IUGR still remains unclear. Diminished cytotrophoblast invasion and uterine spiral artery remodeling leading to placental in- sufficiency have been hypothesized to play crucial roles in the development of IUGR. Hence, the pathomechanisms in IUGR are considered similar to those in preeclampsia by which the mothers are affected. The latter is a new onset hy- pertensive disorder in pregnancy with additional proteinuria, typically occurring after the 20th week of gestation. The as- sumed similarities in pathogenesis of both diseases might explain the fact that both conditions occur simultaneously in C 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com
Transcript
Page 1: Proteome Signature of IUGR From Cord Blood Serum

Electrophoresis 2012, 33, 1881–1893 1881

Manja Wolter1

Claudia Rower1

Cornelia Koy1

Toralf Reimer2

Werner Rath3

Ulrich Pecks3

Michael O. Glocker1

1Proteome Center Rostock,Medical Faculty and NaturalScience Faculty, University ofRostock, Rostock, Germany

2Department of Obstetrics andGynecology, Medical Faculty,University of Rostock, ClinicSuedstadt, Rostock, Germany

3Department of Obstetrics andGynecology, Medical Faculty,RWTH Aachen University,Germany

Received January 1, 2012Revised January 31, 2012Accepted February 16, 2012

Research Article

A proteome signature for intrauterinegrowth restriction derived frommultifactorial analysis of massspectrometry-based cord blood serumprofiling

Intrauterine growth restriction (IUGR) is defined as a condition in which the fetus doesnot reach its genetically given growth potential, resulting in low birth weight. IUGR isan important cause of perinatal morbidity and mortality, thus contributing substantiallyto medically indicated preterm birth in order to prevent fetal death. We subjected um-bilical cord blood serum samples either belonging to the IUGR group (n = 15) or to thecontrol group (n = 15) to fractionation by affinity chromatography using a bead systemwith hydrophobic interaction capabilities. So prepared protein mixtures were analyzed byMALDI-TOF mass spectrometric profiling. The six best differentiating ion signals at m/z8205, m/z 8766, m/z 13 945, m/z 15 129, m/z 15 308, and m/z 16 001 were collectivelyassigned as IUGR proteome signature. Separation confidence of our IUGR proteome sig-nature reached a sensitivity of 0.87 and a specificity of 0.93. Assignment of ion signalsin the mass spectra to specific proteins was substantiated by SDS-PAGE in conjunctionwith peptide mass fingerprint analysis of cord blood serum proteins. One constituent ofthis proteome signature, apolipoprotein C-III0, a derivative lacking glycosylation, has beenfound more abundant in the IUGR cord blood serum samples, irrespective of gestationalage. Hence, we suggest apolipoprotein C-III0 as potential key-marker of the here proposedIUGR proteome signature, as it is a very low-density lipoprotein (VLDL) and high-densitylipoprotein (HDL) member and as such involved in triglyceride metabolism that itself isdiscussed as being of importance in IUGR pathogenesis. Our results indicate that subtlealterations in protein glycosylation need to be considered for improving our understandingof the pathomechanisms in IUGR.

Keywords:

Affinity–mass spectrometry / IUGR / Cord blood analysis / MALDI-TOF MS /Multiparametric profiling / Proteome signature DOI 10.1002/elps.201200001

Correspondence: Dr. Michael O. Glocker, Proteome Center Ros-tock, University of Rostock, Schillingallee 69, D-18057 Rostock,GermanyE-mail: [email protected]: +49-381-494-4932

Abbreviations: AUC, area under curve; CHCA, alpha-cyano-4-hydroxy cinnamic acid; CTRL, control; DHB, 2,5-dihydroxybenzoic acid; HDL, high-density lipoproteins; IUGR,intrauterine growth restriction; [M+H]+, protonated molecu-lar ion; MRM, multi-reaction monitoring; m/z, mass to chargeratio; n-OGP, n-octylglucopyranoside; ROC, receiver operatorcharacteristics; SGA, small for gestational age; TIC, total ioncount; VLDL, very low-density lipoproteins

1 Introduction

Intrauterine growth restriction (IUGR) affects about 3 to 8%of all pregnancies. It is defined as a condition in which thefetus does not reach its genetically given growth potential,resulting in low birth weight. The pathogenesis of IUGRstill remains unclear. Diminished cytotrophoblast invasionand uterine spiral artery remodeling leading to placental in-sufficiency have been hypothesized to play crucial roles inthe development of IUGR. Hence, the pathomechanisms inIUGR are considered similar to those in preeclampsia bywhich the mothers are affected. The latter is a new onset hy-pertensive disorder in pregnancy with additional proteinuria,typically occurring after the 20th week of gestation. The as-sumed similarities in pathogenesis of both diseases mightexplain the fact that both conditions occur simultaneously in

C© 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com

Page 2: Proteome Signature of IUGR From Cord Blood Serum

1882 M. Wolter et al. Electrophoresis 2012, 33, 1881–1893

up to 30% of the cases [1–3]. IUGR is clinically diagnosed by adeceleration of fetal growth velocity in serial antenatal sono-graphic measurements of fetal biometry (crossing centiles)[3, 4]. In clinical practice, one has also considered to defineIUGR as a low estimated fetal weight in addition to signsthat reflect placental insufficiency. Of those, compromisedfetal well being (pathological fetal cardiogram or resistanceindices in Doppler sonographic measurements), asymmetry,and/or reduced amnion fluid index in ultrasonographic ex-aminations have been suggested to be most important [5–9].The clinical challenge for obstetricians is to determine theoptimal time point of preterm delivery in order to reach suffi-cient fetal maturity and preventing fetal death. Hence, IUGRcontributes substantially to medically indicated pretermbirth [3, 5].

Over the last decades, it has become evident that abnor-mal intrauterine conditions also increase the infant’s risk forcardiovascular and metabolic diseases later in life. The as-sociation of being born small for gestational age (SGA) andthe development of atherosclerotic cardiovascular diseaseslater in life has been shown by several epidemiologic studies[10–12]. Such observations have led to the “fetal origins ofdisease hypothesis” or “Barker’s hypothesis” in that abnor-mal conditions in the intrauterine environment may causelong-lasting structural and functional systemic alterations.But by now, exact mechanisms of “fetal programming” arespeculative.

Proteome analyses have the potential to identify markerproteins and/or signatures of proteins of importance for agiven disease state and previous 2D-gel-based analyses ofIUGR cord blood serum samples have shown that inflamma-tory response proteins, immune response proteins, as wellas proteins related to transport, blood pressure, and coagu-lation seem to be of importance [13]. In other 2D-gel-basedstudies, prominent changes of the sialic acid decoration of�2-HS glycoprotein (also termed fetuin-A) were observed inIUGR cord blood serum samples when compared to controls[14] and were discussed as potentially responsible for IUGR-associated complications.

In order to gain deeper insights into protein composi-tions from cord blood samples which were not within reachof hitherto performed investigations but might display infor-mation by which cord blood serum from IUGR samples couldbe characterized and even distinguished from control sam-ples, we have collected cord blood samples from a group ofclinically diagnosed IUGR infants at birth as well as matchedcord blood samples from unaffected newborns. We subjectedall cord blood serum samples to fractionation by affinity chro-matography using a bead system with hydrophobic interac-tion capabilities. So prepared protein mixtures were then an-alyzed by MALDI mass spectrometric profiling, as this ap-proach was previously proven successful in multiparamet-ric characterization of blood samples from pregnant women[15–18]. Since it is evident by epidemiological studies thatIUGR neonates are at increased risk for the development ofcardiovascular and metabolic diseases later in life, we sec-ondary aimed to identify differentially abundant ion signals

in order to detect proteins related to atherosclerosis in um-bilical cord blood, as identifying those proteins could helpdeveloping a targeted therapy for prevention.

2 Materials and methods

2.1 Patient cohorts and controls

The study was approved by the Ethics Committee of theRWTH Aachen (EK 119/08). Written informed consentwas obtained from all participating parents. We analyzedcord blood samples from 30 newborns at the Departmentof Obstetrics and Gynecology, University Hospital of theRWTH Aachen between March 2008 and August 2010. Ofthose, 15 healthy infants were born adequate for gestationalage with 11 being born preterm for various reasons, and15 infants suffered from IUGR with 12 needed manda-tory preterm delivery. Gestational age was established onthe basis of the last menstrual period and confirmed byultrasonic examination between 10th and 14th week ofgestation. Neonatal birth weight centile was determinedaccording to the population-based, newborn weight charts[19]. Sonographic examinations were done antenatally onLogiq 5 R© and Voluson 730 Expert R© Ultrasound Systems (GEHealthcare Systems, Solingen, Germany). The regressionequation including biparietal diameter, femur length, andhead and abdominal circumferences, proposed by Hadlock etal., was used to estimate fetal weight [20]. IUGR was definedin accordance to the ACOG guidelines and as describedrecently [5, 8]. In addition to an estimated fetal weight belowthe 10th percentile, the following criteria had to be fulfilled:(i) deceleration of fetal growth velocity during the last 4weeks, (ii) elevated resistance index in umbilical arteryDoppler sonography above 95th percentile or absent or re-versed end-diastolic blood flow, (iii) fetal asymmetry (head toabdominal circumference ratio above the 95th percentile), or(iv) oligohydramnios (Amniotic Fluid Index < 5 cm). Neona-tal weight was assessed postnatally for correct diagnosis ofSGA < 10th percentile. Additional preeclampsia was definedaccording to the International Society for the Study of Hy-pertension in Pregnancy guidelines [21]. The control group(CTRL) was chosen on the basis to match for gestational ageand gender, and to keep maternal baseline characteristicssimilar. Neonatal weight in the CTRL group was appropriate(within the 10th and 90th percentile). Exclusion criteria werethe following: multiple gestation, fetal anomalies, abnormalfetal karyotype, patients with clinical or biochemical signs ofinfection, positive TORCH screening results, maternal dia-betes mellitus/gestational diabetes, or other severe maternalmetabolic disorders, and patient’s withdrawal from the study.Important clinical characteristics by which mothers and theirneonates were characterized were maternal blood pressureand urinary protein content, week of gestation at delivery,and fetal birth weight centile (Supporting InformationTable S1).

C© 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com

Page 3: Proteome Signature of IUGR From Cord Blood Serum

Electrophoresis 2012, 33, 1881–1893 Proteomics and 2-DE 1883

2.2 Blood sampling, generation, and storage

of serum aliquots

Blood samples (up to 4.9 mL each) were taken postna-tally from the umbilical cord vein using Monovette syringes(Serum 4.9 mL Monovette R©, Sarstedt, Germany). After in-cubation at room temperature for 15–30 min, samples weresubjected to sedimentation of blood cells by centrifugationat 2000 × g at room temperature for 15 min. Serum was as-pirated and divided into aliquots (100 �L each) which werestored at −80�C. Altogether, time between blood sample col-lection and storage of serum aliquots averaged below 1 h.

2.3 Protein extract preparation from serum samples

using the profiling kit 100 MB-HIC8

Serum samples were processed according to established pro-tocols [17] using the Profiling Kit 100 MB-HIC8 (BrukerDaltonik, Bremen, Germany). From each thawed serum sam-ple, 5 �L were incubated with 10 �L MB-HIC8 “bindingbuffer” and 5 �L of MB-HIC8 bead slurry for 1 min. Afterwashing three times (100 �L of “wash buffer”, each), pro-teins were eluted with 10 �L of “elution buffer”, consistingof a 50% ACN solution.

2.4 MALDI-TOF MS profiling of serum proteins and

internal recalibration of mass spectra

After MB-HIC8 extraction, protein solutions (0.5 �L each)were spotted directly onto a stainless steel MTP 384 targetplate (Bruker Daltonik) together with 0.5 �L ferulic acid solu-tion (10 mg ferulic acid dissolved in ACN/0.1% aqueous TFA(33/67, v/v) as matrix. After drying, 0.5 �L ferulic acid solutionwas added to each sample spot again. Protein samples wereprepared in duplicate for recording first and second measure-ments (M1/M2) of the first sample work-up and for recordingfirst and second measurements (M3/M4) of the second sam-ple work-up, respectively. Protein mixtures were analyzedwith a Reflex III MALDI TOF mass spectrometer (BrukerDaltonik) equipped with the SCOUT source and delayed ex-traction and operated in linear positive ion mode using anacceleration voltage of 20 kV [17]. Spectra were recorded ina mass range from 4 to 20 kDa, respectively, accumulating900 shots per spectrum. Spectra were externally calibratedusing a commercially available Protein Calibration Standard(Bruker Daltonik). Furthermore, all mass spectra were in-ternally recalibrated using average masses of ion signals atm/z 13 762.4 (singly charged and unmodified transthyretin;uniprot accession number P02766; and at m/z 6631.6 (singlycharged and unmodified apolipoprotein C, uniprot P02654).Ion signal areas were determined with the ClinProTools 2.2software (Bruker Daltronik) as described previously [18].

2.5 Bioinformatic analysis of linear MALDI-TOF MS

ion signals

The ClinProTools R© 2.2 software (Bruker Daltonik) wasused to analyze the spectra from the patients and con-trols, respectively. Each dataset was generated in dupli-cate, that is, two independent measurements were recordedfrom each sample. The first sample work-up with MB-HIC8 beads resulted in datasets M1 and M2, and the sec-ond work-up in datasets M3 and M4, respectively. Analysisfor each dataset was carried out independently from eachother. Settings have been specified and retained unchangedfor all measurement analyses for (i) peak picking, whichhappened on the total average spectrum of the particular class,(ii) for signal-to-noise threshold, value was set to 5.00, and (iii)the value for the relative threshold base peak was set to 0.00.Smoothing of spectra was performed applying the Savitsky-Golay algorithm, width 2.0 m/z for three cycles. Spectra wererecalibrated automatically, taking only those masses into ac-count which occurred in at least 30% of the spectra. Themaximal mass tolerance between reference mass and peakmass was set to 1500 ppm. Hence, those spectra were ex-cluded from analysis which either could not be recalibrateddue to larger mass differences or which contained too few ionsignals for calibration. Baseline subtraction was performed inthe “Top Hat Baseline mode” and null spectra exclusion wasenabled. All ion intensities in the spectra were normalizedto their own total ion count (TIC), which was determinedas the sum of all intensities of the spectrum. Subsequently,all intensities of this spectrum were divided by the obtainedTIC value. Statistical significance was tested with a Wilcoxonrank-sum test [22] combined with a Kruskal–Wallis test [23],an Anderson–Darling test [24], and an ANOVA test combinedwith a Student’s t-test for two populations [25, 26] on signif-icance levels of 0.05. Graphical representations as box-and-whisker plots [27] were realized using the Origin software(Version 6.1G, OriginLab Corporation, Northampton, MA,USA).

2.6 Protein extract preparation from serum samples

by fractionated precipitation

Serum samples were thawed and fractionated precipita-tion was performed according to described procedures [17].Briefly, to 10 �L of serum sample, 7 �L of chilled ethanol(−20�C; 40% ethanol, v/v) was added and kept on ice for10 min. The pellet (fraction I) was collected by centrifugationwith 13 000 × g at 4�C for 10 min. Proteins dissolved in theice-cold supernatant (17 �L) were precipitated with additionof 3 �L of chilled ethanol (50% ethanol, v/v). After 10 minincubation at −20�C precipitated proteins (fraction II) werecollected by centrifugation as described above. Proteins dis-solved in the ice-cold supernatant (20 �L) were again precip-itated with addition of 5 �L of chilled ethanol (60% ethanol,v/v). The pellet (fraction III) was collected by centrifugationas already described. The dissolved proteins (25 �L) were

C© 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com

Page 4: Proteome Signature of IUGR From Cord Blood Serum

1884 M. Wolter et al. Electrophoresis 2012, 33, 1881–1893

precipitated with addition of 25 �L (80% ethanol, v/v). After10 min incubation at −20�C precipitated proteins (fractionIV) were collected by centrifugation.

2.7 SDS-PAGE and gel imaging

Proteins from the pellet of fraction IV were redissolved in 20�L of SDS sample buffer (2% SDS, 62.5 mM Tris/HCl (pH6.8), 10% glycerol, and bromophenol blue) at room tempera-ture. SDS-PAGE was carried out as described [28]. In short,SDS polyacrylamide gels, 15% T for analyzing proteins inthe low mass range were run for 1 h at 200 V in a Mini-PROTEAN II 2D Cell BIO-RAD Electrophoresis System ap-plying the Laemmli continuous buffer system [29]. Gels werefixed and then stained with colloidal Coomassie Brilliant BlueG-250 [30]. The broad range marker kit (New England Bio-Labs, Frankfurt M., Germany) was used as apparent molec-ular mass calibration. Stained gels were scanned with theUmax Mirage II Scanner (Umax Data Systems, Willich, Ger-many) and images were stored as 16 bit tif files. For den-sitometric image analysis, the software package PhoretixTM

2D Advanced, version 6.01 (Nonlinear Dynamics Ltd,Newcastle upon Tyne, UK) was used.

2.8 Mass spectrometric peptide mass fingerprinting

Protein bands of interest were excised manually and gel plugswere subjected to in-gel digestion with trypsin (Promega,Mannheim, Germany) [16]. Sample preparation of peptidemixtures was performed on an AnchorChipTM 600/384 tar-get plate [28] using DHB as matrix. Peptide mixtures wereanalyzed with a Reflex III MALDI TOF mass spectrometer(Bruker Daltonik) equipped with the SCOUT source and de-layed extraction and operated in positive ion mode using anacceleration voltage of 20 kV. Spectra were externally cal-ibrated using a commercially available Peptide CalibrationStandard (Bruker Daltonik) as well as internally recalibratedusing the following peptide ion signals derived from trypsinautoproteolysis: [M+H]+ 842.51, [M+H]+ 1045.54, [M+H]+

2211.10, [M+H]+ 2807.39. Mass spectra were further pro-cessed and analyzed with the FlexAnalysis 2.4 and BioTools3.0 softwares. Database searches were performed against anin-house SWALL database that consists of Swiss-Prot release51. and TrEMBL release 34.0 using Mascot version 2.2.07 soft-ware (Matrix Science, London, UK) with the following searchparameters: taxonomy: human, peptide tolerance: 60 ppm,fixed modifications: carbamidomethylation of cysteines, vari-able modifications: oxidation of methionines, one missedcleavage site.

2.9 Bioinformatic and biostatistical analysis

of mass spectra

Data analysis was carried out using the statistical packagesSAS Version 9.1 Software (SAS Institute, Cary, NC, USA).Clinical data were evaluated by two-way ANOVA (analysis of

Figure 1. MALDI TOF mass spectrum of a serum sample af-ter ClinProt R© work-up from an IUGR cord blood serum sample.Mass range m/z 6000–20 000. The inserts show zooms into themass ranges m/z 8500–9400 and m/z 13 700–14 000, respectively.* indicates ion signals that were used for internal recalibration;ferulic acid was used as matrix. For assignments of protein ionsignals, see Supporting Information Table S2.

variance) and expressed as mean and 95% confidence interval(CI). Differences of serum parameters were tested for signif-icance using Mann–Whitney test. Association analyses wereperformed using Spearman’s coefficient of rank correlation(rho). Sensitivity, specificity, and likelihood ratios were mea-sured on receiver operator characteristic (ROC) plots usingthe Origin software (Version 6.1G, OriginLab Corporation)and ClinProTools R© 2.2 software.

Gel views of mass spectrometric data were generated us-ing an in house visualization tool based on a PHP script(Linux Command Line Interpreter) [15]. The ion signal inten-sity threshold was set to 3500. Values below 3500 are encodedin linear gray scales from 0 (baseline; white) to 255 (maxi-mum; black). Spectra (m/z range 8500–9050) were importedas character separated csv files as exported from Flex Analysis(Bruker Daltonik) after baseline correction.

Hierarchical clustering was performed based on the com-plete linkage method and Spearman’s correlation coefficientas a measure of similarity. Signal intensities were centeredand scaled row-wise for visualization purposes [31]. Normal-ized values of selected spots are graphically represented inheat maps.

In order to evaluate the minimal required sample sizethat is needed to discriminate the group with IUGR from thecontrol group on the basis of the obtained data, we carriedout a power analysis [32] using the GPower 3.0.10 statisticalsoftware. We chose a type I error (�) of 0.05 and a type IIerror (�) of 0.20 in a comparison of two means which istypically suggested for this type of analysis. Thus, we achieveda minimal required sample size with a power (1-�) of 80%and a level of significance below 0.05.

2.10 Mass spectrometric peptide sequencing

Peptide mixtures from in-gel digests (1 �L) were mixedwith matrix solution (1 �L) on an AnchorChipTM 600/384

C© 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com

Page 5: Proteome Signature of IUGR From Cord Blood Serum

Electrophoresis 2012, 33, 1881–1893 Proteomics and 2-DE 1885

Figure 2. Virtual gel view of MALDI mass spectra in “gray scale”mode. Mass range m/z 8500–9000. Spectra were recorded in du-plicate from healthy control cord blood serum samples (rows 1–15) and from IUGR samples (rows 16–30). After background sub-traction and normalization, the baseline is represented in whitecolor (intensity 0) and saturated ion signals in black color (inten-sity 1). The region around the ion signal with m/z 8766 is boxed.Patient numbering as in Supporting Information Table S1.

target plate. The matrix solution contained 5 mg/mL 2,5-dihydroxybenzoic acid (DHB) which were dissolved in 1 mLof solvent that consisted of 50% ACN and 50% of a 0.1%TFA solution (v/v). MS/MS spectra were recorded on an Ax-ima MALDI QIT ToF mass spectrometer (Shimadzu Biotech,Manchester, UK) utilizing a nitrogen pulsed laser (337 nm,3–5 ns pulse length) and employing a three-dimensionalion trap supplied by helium (pulsed flow gas) for colli-sional cooling and argon (collisional gas) for collisionallyinduced dissociation (CID) [33, 34]. Precursor ions were ex-cited with off-resonance sinusoidal waveforms whereas atthe same time argon was allowed to enter the trap. Forfragmentation, the width of the precursor ion selection win-dow was dependent on the mass of the precursor ion and

Figure 3. Distribution of quotient values for selected ion intensitydifferences from measurements 1 and 2 after ClinProt R© work-upof cord blood serum samples depicted as box and whisker plots.(A) m/z 8205 and m/z 8766. (B) m/z 8766 and m/z 13 945. (C) m/z8766 and m/z 9422. (D) m/z 16 001 and m/z 13 883. (E) m/z 15308 and m/z 13 883. (F) m/z 15 129 and m/z 13 883. The boxesrepresent the 25th–75th percentiles. The horizontal lines withinthe boxes represent the median, the small squares indicate themean. The whiskers specify the 5th and 95th percentiles, and thecrosses indicate the 1st and 99th percentiles. Dashed lines markselected cut-off values. CTRL, control.

varied between 2 and 10 Da. Spectra were externally cali-brated with a manually mixed peptide standard consistingof bradykinin (1–7), [M+H]+ 757.39; angiotensin II, [M+H]+

1046.53; angiotensin I, [M+H]+ 1296.68; bombesin, [M+H]+

1619.81; N-acetyl renin substrate, [M+H]+ 1800.93; ACTH(1–17), [M+H]+ 2093.08; ACTH (18–39), [M+H]+ 2465.19;somatostatin, [M+H]+ 3147.46; insulin (oxidized beta chain),[M+H]+ 3494.64; as well as internally recalibrated usingknown ion signals of the MS/MS spectrum. Further process-ing and analysis of the MS/MS spectra was performed withthe LaunchpadTM software, version 2.8.4 (Shimadzu Biotech).

2.11 ELISA assay for apolipoprotein C-III

A commercially available ELISA test kit (AssayMax HumanApoC-III ELISA Kit, St. Charles, MO, USA) was used in or-der to quantitatively analyze ApoC-III expression in sera. Allmeasurements were performed in duplicate. The absorbancewas measured at 450 nm, using the Rosys Anthos 2010 in-strument (Anthos Labtec Instruments, Austria).

C© 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com

Page 6: Proteome Signature of IUGR From Cord Blood Serum

1886 M. Wolter et al. Electrophoresis 2012, 33, 1881–1893

Figure 4. Hierarchical clustering analysis with quotient values de-rived from protein ion signal intensities. Duplicate measurementsfrom control samples are clearly sorted to the CTRL group (left)and IUGR samples to the IUGR group (right), except for controlsamples 14 (one sample) and 15 (both samples) that are allocatedto the IUGR group, and IUGR samples 16 (both samples), 17 (onesample), and 18 (one sample) that are grouped to the controls.Quotients A–F from top to bottom as in Fig. 3 (indicated at theright). Patient numbering as in Supporting Information Table S1.

3 Results

3.1 Clinical characteristics of patients and donors

Clinical parameters by which the study group (IUGR) and thecontrol group (CTRL) could be characterized include both, in-formation about the fetus/infant and the mothers. Importantfacts on the mothers are maternal age (IUGR: 30.4 years(95% CI 26.9–34.0), CTRL: 31.7 years (95% CI 28.3–33.9)),maternal BMI (IUGR: 24.2 (95% CI 22.4–26.1), CTRL: 23.7(95% CI 20.8–26.6)), primipariety (IUGR: 80%, CTRL 27%),and smoking status of the mother (currently smoking IUGR:27%, CTRL 20%). The mean week of gestation at deliverywas 33.0 wk (95% CI 30.5–35.5) for the IUGR group, and32.8 wk (95% CI 30.2–35.3) for the CTRL group (SupportingInformation Table S1). Mean fetal weight was 1383 g (95%CI 1005–1761) in IUGR and 2129 g (95% CI 1584–2673)for CTRL. By definition, IUGR fetuses were born with birthweight below the 10th percentile (IUGR: mean 4.0 (95% CI2.5–5.5), CTRL: 50.7 (95% CI 40.8–60.5)).

Blood samples were taken postnatally from the umbilicalcord vein and serum was prepared and divided into aliquotsaccording to standardized protocols. Altogether, time be-tween blood sample collection and storage of serum aliquotsaveraged below 1 h.

3.2 Serum profiling by MALDI TOF MS analysis and

patient screening

All cord blood serum samples were individually subjected toprotein fractionation using hydrophobic bead surfaces. Thisprocedure proved very efficient to remove highly abundantserum protein species, such as albumin and immunoglob-ulins that together make up more than 70% of the serumprotein content. The resulting mixture was reflecting serumproteins (i) of moderate abundance, (ii) with strong bindingaffinities to the bead surface, and (iii) which could be eluted

using volatile aqueous buffer systems used for preparation ofthe samples on MALDI-MS targets.

MALDI-MS measurements were recorded in duplicate,that is, from one preparation two spectra were collected foreach sample, generating a set of 60 spectra (series MS1 andMS2). A typical mass spectrum showed more than 60 ionsignals that were reproducibly recorded in the mass rangebetween m/z 4000 and m/z 25 000 (Fig. 1), forming twogroups of intense ion signals. On average signal-to-noise ra-tios of 10:1 and better were recorded for small ion signals(e.g. the lowest S/N in the IUGR group of 6:1 was observedat m/z 8766 and at m/z 13 945, respectively).

The first group of ion signals was observed in the massrange m/z 6000 and m/z 10 000 correlating to predominantlydoubly charged (protonated) proteins, such as transthyretinand its modifications (Supporting Information Table S2). Butin this mass range also singly charged ion signals of smallproteins (e.g. apolipoprotein C-III) were observed.

The second group of ion signals was present in the massrange of m/z 13 000 to m/z 18 000 containing predominantlysingly charged ion signals. Hemoglobins and apolipoproteinA-II are prominent examples of the proteins that were causingthe observed ion signals (Supporting Information Table S2).Protein assignments of ion signals were made possible bycomparison with literature references [17, 35–40].

As all spectra showed high similarities to each other,semi-quantitative differential analysis of ion signal intensitieswas found feasible using a statistical approach. According tothe calculated p-values, the best differentiating singly chargedion signals between the two sample groups were found at m/z8205, m/z 8766, m/z 13 945, m/z 15 129, m/z 15 308, and m/z16 001 (Table 1).

From these ion signals as well as from two ion signalsat m/z 9422 and m/z 13 883 which were found reproduciblypresent in all spectra (so-called landmark signals), the areasunder the ion signals were determined (Supporting Informa-tion Table S3). In order to reproduce the results, a secondpreparation was conducted, leading to a second set of mea-surements (series MS3 and MS4; Supporting InformationTable S4). This second set of spectra was later used to applythe discrimination rules that were determined using the firstset of data (see below).

3.3 Test accuracy estimation upon multiparametric

analysis

Because of the close similarities of the spectra, that is, numberof ion signals and relative intensities of them, independentof their clinical grouping into either the CTRL group or theIUGR group, ion signal areas that were differentiating be-tween the two groups were also of low intensity in all spectra(Fig. 2). By using a “gel view” image, even small but clearlypresent differences in intensities (areas) of small ion signalscan be visualized in the vicinity of large (nondifferentiating)ion signals by distinct gray values that represent the respec-tive intensities (area values). The ion signal at m/z 8766, for

C© 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com

Page 7: Proteome Signature of IUGR From Cord Blood Serum

Electrophoresis 2012, 33, 1881–1893 Proteomics and 2-DE 1887

Table 1. Statistical data of ion signals selected for multiparametric analysis

Entrya) m/z PTTAb) PWKWb) PADb) DAvec) Peak area [a.u.]/standard deviation

CTRL IUGR

1 8205 0.0000155 <0.000001 0.00000845 201.55 368.66/168.80 167.11/61.732 8766 0.000138 0.0000188 0.534 28.06 48.58/14.17 76.64/28.703 13945 0.000138 0.0000775 0.95 29.13 74.94/20.14 45.81/27.344 15129 0.0000331 <0.000001 <0.000001 830.61 597.34/264.16 1427.95/759.195 15308 0.0000155 <0.000001 <0.000001 233.72 165.12/64.22 398.84/199.506 16001 0.00321 0.00000437 <0.000001 611.11 394.64/242.18 1005.74/895.927 9422 0.00105 0.00122 0.18 244.12 721.47/249.24 477.34/237.468 13883 0.00364 0.000191 0.183 33.6 98.64/29.63 65.05/44.79

a) Entries 1–6: best differentiating ion signals according to peak statistics; entries 7, 8: most robust ion signals according to peak areas.b) PTTA, p value from combined paired Student’s t-test and ANOVA test; PWKW, p value from combined Wilcoxon rank-sum test andKruskal–Wallis test; PAD, p value from Anderson–Darling test.c) DAve, difference between the maximal and the minimal average peak area/intensity of all classes.

Figure 5. SDS-PAGE analysis of IUGR cord blood serum proteins.Protein bands (colloidal Coomassie blue staining) after fraction-ated precipitation (lane 2) are shown. Mass spectrometricallyidentified proteins in the bands (numbered 1–11 at the right) aregiven in Table 4. The locations of marker proteins with knownapparent masses are given at the left (lane 1).

example, was consistently found more intense in the serumsamples from the IUGR group with respect to the serumsamples of the controls.

However, the area value distributions of just one ionsignal were not discriminating enough to convincingly sort

a given spectrum into either the IUGR or the CTRL group.By contrast, it turned out that the two most robust singlycharged ion signals and the six best differentiating singlycharged ion signals carried enough information for accuratesorting of individual spectra, that is, of individual samples,after applying rules by which the relations of the respectiveion signal intensities were considered.

The simplest rules that we applied were to relate theselected ion signals pairwise, forming quotients of the areasof the respective ion signals at (A) m/z 8205 and m/z 8766,(B) m/z 8766 and m/z 13 945, (C) m/z 8766 and m/z 8422, (D)m/z 16 001 and m/z 13 883, (E) m/z 15 308 and m/z 13 883,as well as (F) m/z 15 129 and m/z 13 883. The resultingdistributions of the ratios were determined for each of thesix pairs (Fig. 3) and showed in fact significant differences(p-values below 0.01) in all six cases.

Next, cut-off values (dashed lines in Fig. 3) were deter-mined for all six signal intensity ratios such that they werelocated between the upper 75% quartile of the group wherelower values were obtained (CTRL group) and below the 25%quartile of the other group (IUGR) where on average highervalues were calculated for the respective ratios of ion signalareas. Then, it was tested whether in a given sample spec-trum the respective cut-off value was reached or not. Whenthe value of one quotient was higher than the respective cutoff, a score of “1” was given to this respective sample. In thecontrary case, the score for this sample was set to “0”. Thischeck was carried out for each of the six ion signal ratiosindependently. Hence, each spectrum, that is, each sample,could ultimately reach a cumulative score between “0” and“6”.

Finally, it was determined that a cumulative score above“2” was sorting the respective spectrum (sample) into theIUGR group. Applying these grouping rules for the val-ues of the first measurement series (MS1 and MS2) en-abled a clear-cut separation of IUGR samples into theIUGR group and all control samples into the CTRL group(Table 2).

C© 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com

Page 8: Proteome Signature of IUGR From Cord Blood Serum

1888 M. Wolter et al. Electrophoresis 2012, 33, 1881–1893

Table 2. Statistical significance of mass spectrometric profiling results

Measurement True False True False Sensitivity Specificity False False Positive Negative ROCpositive positive negative negative positive rate negative rate predictive value predictive value AUC c)

MS 1+2a) 28 1 29 2 0.93 0.97 0.03 0.07 0.97 0.94 0.99MS 3+4b) 26 2 28 4 0.87 0.93 0.07 0.13 0.93 0.88 0.91

a) Values from the first measurement series (MS1 and MS2) of the first sample work-up.b) Values from the second measurement series (MS3 and MS4) of the second sample work-up.c) ROC, receiver operator characteristics; AUC, area under curve.

Figure 6. MALDI-QIT-TOF MSn sequencing result after fragmen-tation of peptide ion signal at m/z 1196.61 Determined partialamino acid sequences are depicted and were assigned to theapolipoprotein C-III-derived peptide comprising amino acids 41–51. Amino acid residues are depicted in single letter code. Themass spectrometric fragment ions from the B-type ion series, theY’’ ion series, as well as typical losses of water (*) are indicated.#, not identified. DHB was used as matrix.

Ultimately, the scoring system resulted in a sensitivity of0.93, a specificity of 0.97, a false positive rate of 0.03, a falsenegative rate of 0.07, a positive predictive value of 0.97, and anegative predictive value of 0.94 for the measurement set usedfor determining the separation rules (retrospective analysis).Similar results were obtained for the test set (measurementseries MS3 and MS4; prospective analysis) applying the ex-act same rules for spectra classification but using the newlydetermined area values. With both preparations, the AUCvalues of the ROC characteristics of either measurement se-ries were better than 0.9, which is considered satisfactory forclinical purposes.

All six quotient value distributions A–F were also usedfor hierarchical clustering of the individual samples from thefirst preparation (Fig. 4) which resulted in clear separation ofthe samples according to their clinical classification.

Samples located at the left of the hierarchical tree werebelonging to the controls, except for the two measurementsfrom samples 16. In addition, only one measurement fromsample 17 and one from sample 18 was assigned to the con-trol group although clinically assigned as being IUGR sam-ples. Accordingly, samples located to the right were assignedas IUGR samples with the exception of both measurementsfrom sample 15 and one measurement from sample 14 that

both were controls according to clinical data (cf. Support-ing Information Table S1). Interestingly, the IUGR groupseemed much more homogenous except for samples 29 and30 (at the very right in Fig. 4). These samples obviously con-tained too much hemoglobin as judged by their reddish color,presumably because of artifactual disruption of red blood cellsduring serum preparation. Therefore, these two samples wereexcluded from subsequent subgroup analyses. As indicated,the control group seemed more heterogeneous, consistingof three major groups. Higher heterogeneities in the controlgroups were also observed in earlier studies [17] and is at-tributed to the fact that in general clinical inclusion criteriaare well defined for the disease under study but less welldefined for controls (absence of disease).

3.4 Subgroup analysis of ion signal intensity

differences with respect to gestational age

When analyzing cord blood protein compositions, one hasto consider that patients suffering from severe IUGR are of-ten delivered preterm. Thus, the careful selection of a controlgroup matched for gestational age at delivery is required toavoid confounding, because prematurity per se is a pathologiccondition and serum parameters might be biased by factorsleading to premature labor. We therefore tested whether theprotein ion signals that were included into our marker sig-nature for IUGR were affected by gestational age and splitpatient samples (cf. Supporting Information Table S1) intotwo groups, that is, samples that were obtained from indi-viduals with less than 34 weeks of gestation were separatedfrom those after 34 weeks of gestation. Next, the mean val-ues of intensities from our signature ion signals and alsothe differences of the means were compared to each other(Table 3).

Intensity differences of the ion signals at m/z 8205, m/z13 883, m/z 13 945, m/z 15 129, and m/z 15 308 increasedwith gestational age. The one ion signal whose differencesin intensities between the IUGR group and the controls di-minished in the “late” samples (i.e. more than 34 weeks ofgestation) as compared to the “early” stages was that with m/z16 001. Interestingly, the only ion signal whose intensity dif-ference was not significantly affected by gestational age wasthat at m/z 8766. Hence, the protein that originated this ionsignal was considered as being of further interest.

C© 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com

Page 9: Proteome Signature of IUGR From Cord Blood Serum

Electrophoresis 2012, 33, 1881–1893 Proteomics and 2-DE 1889

Table 3. Subgroup analysis of differentiating ion signal intensities from cord blood samples with respect to gestational age

Ion signal [m/z] Gestation < 34 weeksa) Gestation > 34 weeksa) Differencesb)

CTRL IUGR CTRL IUGR <34 weeks >34 weeks

8205 311.68 161.62 454.14 175.36 +150.06 +278.788766 48.18 77.82 49.16 74.87 −29.64 −25.7113 883 93.49 74.28 106.38 51.19 +19.21 +55.1913 945 70.20 54.04 82.05 47.65 +16.16 +34.4015 129 668.38 1204.74 490.78 1326.81 −536.36 −836.0315 308 182.45 350.01 139.11 369.48 −167.56 −230.3716 001 470.12 707.27 281.41 397.74 −237.15 −116.33

a) Mean values of ion signal intensities from measurement series MS1 and MS2 (cf. Supporting Information Table 2).b) CTRL value minus IUGR value.

Table 4. Identified proteins in SDS-PAGE analysis after fractionated precipitation of serum proteins from cord blood

Band Accession Protein name Scorec) RMS Matchede) Coverage Molecularnumbera) numberb) [ppm]d) [%]f) mass [Da]g)

1 P02765 Alpha-2-HS-glycoprotein 93 37 8 22 40 0982 P06727 Apolipoprotein A-IV 377 26 30 62 45 3713 P02766 Transthyretin 92 53 5 55 15 9914 P02647 Apolipoprotein A-I 246 17 23 64 30 7595 P69891 Hemoglobin subunit gamma-1 172 29 11 74 16 187

P69892 Hemoglobin subunit gamma-2 172 29 11 74 16 1736 O95445 Apolipoprotein M 158 34 11 35 21 5827 P02753 Retinol-binding protein 4 127 23 11 68 23 3378 P02652 Apolipoprotein A-II 100 24 6 64 11 2829 P02766 Transthyretin 197 34 9 69 15 99110 P69891 Hemoglobin subunit gamma-1 158 19 10 74 16 187

P69892 Hemoglobin subunit gamma-2 158 19 10 74 16 17311 P02656 Apolipoprotein C-III h) 48 25 4 51 10 845

a) Band numbers refer to gel in Fig. 5.b) Uniprot data base entry.c) Probability based MOWSE score.d) Root mean square error.e) Number of ion signals that match with database entry.f) Sequence coverage of ion signals with respect to database entry.g) Calculated from database entry.h) Confirmed by MALDI-QIT-TOF-MS/MS sequence analysis (cf. Fig. 6).

3.5 Evaluating apolipoprotein C-III results by

SDS-PAGE and ELISA

In order to substantiate protein assignments of the serumsamples, we carried out an SDS-PAGE analysis after fraction-ated precipitation of serum proteins (Fig. 5). Fractionatedprecipitation eases SDS-PAGE analysis of low abundant pro-teins but because of difficulty to reproduce precipitation con-ditions is not considered to be precise enough to be applicablefor differential quantitative analysis. In addition, separationpower of SDS-PAGE is not sufficient to guarantee avoidanceof comigration of serum proteins. About 11 strongly stainedbands were visible in the gel with the protein mixture in themass range between 55 kDa and 6 kDa in both, the IUGRsamples and in the control samples (lane 2 shows the proteindistribution of a representative patient sample). Proteins inthese intense bands were subjected to mass spectrometricidentifications after in-gel digestion with trypsin.

We found that band 11, migrating at 9 kDa apparentmass, contained apolipoprotein C-III (P02656) as the ma-jor compound (Table 4); although with low identificationscore. Band 4 was determined to contain apolipoprotein A-I (P02647). Band 2 contained apolipoprotein A-IV (P06727),and band 1 at ca. 53 kDa apparent mass contained alpha-2-HS-glycoprotein (P02765; fetuin A). With the exceptionof apolipoprotein C-III in band 11, all other protein iden-tifications were obtained with good scores (above 90), in-dicating that the identified proteins were in fact the majorcompounds in their respective bands. However, differentiat-ing between hemoglobin subunit gamma 1 and hemoglobinsubunit gamma 2 in bands 5 and 10, respectively, was notpossible.

In order to confirm the identification result forapolipoprotein C-III from band 11, we performed mass spec-trometric peptide sequencing of the ion with m/z 1196.61using a MALDI-QIT-TOF MSn instrument. The MS/MS

C© 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com

Page 10: Proteome Signature of IUGR From Cord Blood Serum

1890 M. Wolter et al. Electrophoresis 2012, 33, 1881–1893

spectrum (Fig. 6) showed a dominating ion signal at m/z638.37 which can be explained by cleavage of a D-G bond(Y′′

6 fragment) that is present in the suspected amino acidsequence comprising amino acids 41–51 (sequence: GWVT-DGFSSLK). Due to unique instrument specificities, furthersequence-specific fragment ion signals, however with muchlower intensities, were generated in the QIT ion trap togetherwith internal fragments and loss of water ion signals, con-firming the sequence of peptide 41–51 from apolipoproteinC-III.

As the ion with m/z 1196.61 is reported to belong toa proteotypic peptide used for MRM assays [41], the identi-fication of apolipoprotein C-III was considered solid. As noadditional proteins could be identified in this band, it seemedlikely that the major protein component in this band was infact apolipoprotein C-III.

Since the SDS-PAGE-derived protein identification re-sults were in agreement with the MALDI-MS-based assign-ments, and since the ion signal intensity differences of thesignal at m/z 8766, assigned to apolipoprotein C-III0 weredistinct but not affected by gestational age, we initiatedquantification experiments for apolipoprotein C-III by ELISAusing a commercially available assay (Table 5).

In ELISA analyses, the apolipoprotein C-III concentra-tions in cord blood samples varied approximately by a factorof seven between individuals, ranging from 10.1 �g/mL to69.5 �g/mL. Although there seemed to be a trend for de-creased apolipoprotein C-III levels in the IUGR group, dis-tinctly different apolipoprotein C-III concentrations betweenIUGR samples and controls were not evident (p > 0.05).

Obviously, the mass spectrometrically observed differ-ences between the ion signal intensities at m/z 8766 whichwere attributed to belonging to apolipoprotein C-III0, werenot present in ELISA analyses. This lack of correlation may beexplained by the fact that this ELISA assay does not differen-tiate between apolipoprotein C-III sialylation forms. Hence,we estimated the relative abundances of all four apolipopro-tein C-III-derived ion signals (m/z 8766 (Apo C-III0), m/z9134 (Apo C-III0

′), m/z 9422 (Apo C-III1), and m/z 9713 (ApoC-III2); Table 5), by summing up their ion signal areas to100%. In this apolipoprotein C-III-focused investigation weagain found that the best differentiating protein species be-tween samples from the IUGR group and the controls wasapolipoprotein C-III0. Whereas in the CTRL group, the rela-tive amount of apolipoprotein C-III0 was around 3–4% (withexception of sample 15) of the total apolipoprotein C-III, therelative amount of apolipoprotein C-III0 in the IUGR groupwas around 8–10%.

4 Discussion

We recently reported on affinity-based, multiparametricMALDI-TOF MS analysis of blood serum samples as an over-all robust and valid method for both, biomarker discovery,and for diagnostic purposes in pregnancy-related disorders[17, 18]. We now adapted this method for the analysis of um-

bilical cord blood samples by focusing on neonates that wereborn after pregnancies were complicated by IUGR.

To our knowledge, two studies using proteomic ap-proaches for the analysis of umbilical cord blood comparingIUGR and control groups have been reported [13,14]. In both,differential 2D gel electrophoresis had been performed fol-lowed by peptide mass fingerprinting to identify the proteinsof interest. In the first study, 18 protein spots were reportedto be differentially expressed in umbilical cord serum sam-ples as compared to controls. The identified proteins fromthese spots were discussed as being involved in the coagula-tion process, immune mechanisms, blood pressure regula-tion, and iron or copper homeostasis [13]. Unfortunately, theauthors did not provide sufficient information about neitheradditional antenatal sonographic examinations to confirm thediagnosis IUGR nor to the week at delivery. According to ourexperience, IUGR fetuses are born significantly earlier thannormal and, if compared to a full-term control group, detectedprotein alterations in cord blood serum might be overlayedby protein composition changes that are due to the gesta-tional age of the fetus at delivery. The other study focusedon late-onset IUGR where infants were born round aboutafter 39 weeks of gestation. Here, 16 proteins were reportedto be differentially expressed. The most interesting findingof this study was that the negative acute phase parameter �2-HS glycoprotein (fetuin-A) was decreased in IUGR samples.Interestingly, the authors found that in the IUGR samplesfetuin-A forms, lacking O-linked sialic acid residues, wereuniquely present while absent in plasma samples of the con-trol group [14].

In contrast to these two published studies, we here fo-cused on cord blood serum proteins of low mass range whichusually escape detection by 2D gel electrophoresis but can beinterrogated by affinity-based MALDI-TOF MS as a screen-ing tool. Our results show that protein abundance differencesbetween IUGR cord blood samples and CTRL samples canbe reliably determined using an affinity-enrichment proce-dure combined with MALDI-MS profiling. Even more, massspectrometry-based multiparametric analysis of cord bloodsamples is capable of differentiating individual samples fromthe IUGR group from those of the control group with highconfidence.

We recently hypothesized that alterations in cord bloodcomposition toward an atherogenic phenotype secondary toplacental insufficiency may underlay at least in part the pre-disposition of the development of cardiovascular diseases inIUGR [42] as we found atherogenic lipoprotein indices andtotal triglyceride concentrations to be significantly increasedin IUGR as compared to controls. In this context it is of inter-est that we found in this study that apolipoprotein C-III0 wasthe most prominent apolipoprotein C-III derivative with dif-ferential abundance. Apolipoprotein C-III is a 79 amino acidconstituent of both apo B- and apo A-I-containing lipopro-teins [43, 44]. In normolipidemic individuals apolipoproteinC-III is found in HDL and VLDL particles whereas in pa-tients with elevated plasma triglycerides it is associated pre-dominantly with triglyceride rich lipoproteins associated with

C© 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com

Page 11: Proteome Signature of IUGR From Cord Blood Serum

Electrophoresis 2012, 33, 1881–1893 Proteomics and 2-DE 1891

Table 5. Total apolipoprotein C-III concentrations and relative amounts of sialylated species in cord blood samples

Patient Apo C-III Standard Apo C-III0 Apo C-III0´ Apo C-III1 Apo C-III2

[�g/mL]a) deviation [%]b) [%]b) [%]b) [%]b)

MS 1 MS 2 MS 1 MS 2 MS 1 MS 2 MS 1 MS 2

1 54.68 ±0.060 4 3 5 5 49 49 42 432 69.49 ±0.018 4 3 5 6 56 57 35 343 33.21 ±0.042 4 3 5 5 60 60 31 324 16.96 ±0.028 5 4 5 5 64 61 26 305 21.94 ±0.004 2 3 6 5 55 54 37 386 43.33 ±0.032 2 2 6 7 60 59 32 327 26.55 ±0.043 3 4 8 7 56 56 33 338 15.46 ±0.013 5 8 8 7 58 58 29 279 22.04 ±0.006 2 2 10 10 57 57 31 3110 25.05 ±0.015 5 5 7 8 58 55 30 3211 22.44 ±0.016 7 8 8 8 53 56 32 2812 27.43 ±0.014 6 5 5 5 56 59 33 3113 18.79 ±0.016 4 4 6 6 58 56 32 3414 35.98 ±0.025 2 2 4 4 57 54 37 4015 15.17 ±0.023 8 9 6 8 57 56 29 2716 15.94 ±0.005 9 8 8 7 49 50 34 3517 26.01 ±0.029 7 9 6 5 58 56 29 3018 10.11 ±0.000 9 8 4 5 43 40 44 4719 25.93 ±0.003 5 7 6 7 60 58 29 2820 12.99 ±0.011 9 8 4 4 50 49 37 3921 24.74 ±0.018 9 9 3 4 44 44 44 4322 18.98 ±0.004 8 7 5 5 46 46 41 4223 46.64 ±0.016 7 6 6 6 65 63 22 2524 14.96 ±0.008 10 9 5 5 48 47 37 3925 26.87 ±0.015 11 15 5 6 55 49 29 3026 10.32 ±0.001 11 10 10 11 55 53 24 2627 34.43 ±0.021 6 5 4 4 50 50 40 4128 20.93 ±0.018 9 11 5 6 50 50 36 3329 34.11 ±0.045 10 9 11 11 67 68 12 1230 15.12 ±0.006 21 18 7 9 43 49 29 24

a) ELISA determinations.b) Relative peak areas of ion signals at m/z 8766 (Apo C-III0), m/z 9134 (Apo C-III0 ′), m/z 9422 (Apo C-III1), and m/z 9713 (Apo C-III2),respectively.

LDL and VLDL particles [45]. Under physiological conditions,apolipoprotein C-III exists in four major species designatedas C-III0, C-III0

′, C-III1, and C-III2, where the subscript indi-cates the number of sialic acid residues. In previous studies,it was found that apolipoprotein C-III2 binding to VLDL was2-fold greater as compared with C-III0 and C-III1 binding,and simultaneously displayed a decreased inhibitory effecton the lipolysis stimulated receptor (LSR) that is involved inthe clearance of triglyceride-rich proteins [46]. From theseobservations and from animal experiments apolipoproteinC-III0 has been suggested to delay triglyceride catabolismlinking it to hypertriglyceridemia and to atherosclerotic dis-eases [47, 48]. Moreover, it was reported that loss of sialicacid residues from lipoprotein particles in general can in-duce cholesterol-ester accumulation in human aortic smoothmuscle cells. Thus, desialylation of lipoproteins was con-sidered directly involved in the early stages of atherogene-sis characterized by foam cell formation [49]. The observa-tion of desialylated apolipoprotein C-III0 being increased in

IUGR samples stands in agreement to these observationsand might provide at least a partial explanation for the in-creased triglyceride levels in cord blood serum. Hence, thisassumed association between apolipoprotein C-III0 proteinabundance, and triglyceride levels in cord blood samples isworth to be analyzed further and is current focus of ongoingstudies.

Even more, since another protein, fetuin-A, was presentin its desialylated form in IUGR cord blood samples as well[14], it is tempting to speculate whether enzymatic desialyla-tion and/or release of immaturely sialylated protein deriva-tives into circulation might be regarded as a common phe-nomenon in the pathogenesis of IUGR. Protein-bound sialicacid residues are involved in a variety of physiological andpathological processes including modulation of fertilizationand development, alteration of immune response [50], on-cologic diseases, and brain development (for reviews see[51, 52]). Of particular interest is the finding that geneticelimination of sialic acid production in the mouse results in

C© 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com

Page 12: Proteome Signature of IUGR From Cord Blood Serum

1892 M. Wolter et al. Electrophoresis 2012, 33, 1881–1893

embryonic lethality [53]. Thus, studying alterations in sialicacid metabolism seems likely to provide a novel approachfor improving our understanding of the pathomechanismsin IUGR.

This work was supported by grants from the University ofRostock and the University of Aachen. The authors like to ex-press their thanks to Mrs M. Ruß, Mr. M. Kreutzer, and Mr. F.Steinbeck for valuable methodological help.

The authors have declared no conflict of interest.

5 References

[1] Huppertz, B., Hypertension 2008, 51, 970–975.

[2] Steegers, E. A. P., von Dadelszen, P., Duvekot, J. J.,Pijnenborg, R., The Lancet 2010, 376, 631–644.

[3] Villar, J., Carroli, G., Wojdyla, D., Abalos, E., Gior-dano, D., Ba’aqeel, H., Farnot, U., Bergsjø, P., Bakketeig,L., Lumbiganon, P., Campodonico, L., Al-Mazrou, Y.,Lindheimer, M., Kramer, M., Am. J. Obstet. Gynecol.2006, 194, 921–931.

[4] ACOG practice bulletin. Int. J. Gynaecol. Obstet.: TheOfficial Organ of the International Federation of Gynae-cology and Obstetrics 2000, 68, 175–185.

[5] Intrauterine growth restriction. Int. J. Gynaecol. Obstet.:The Official Organ of the International Federation of Gy-naecology and Obstetrics 2001, 72, 85–96.

[6] al Riyami, N., Walker, M. G., Proctor, L. K., Yinon, Y.,Windrim, R. C., Kingdom, J. C. P., J. Obstet. Gynaecol.Canada 2011, 33, 715–719.

[7] Campbell, S., Thoms, A., Brit. J. Obstet. Gynecol. 1977,84, 165–174.

[8] Pecks, U., Caspers, R., Schiessl, B., Bauerschlag, D.,Piroth, D., Maass, N., Rath, W., Hypertens. Pregnancy2012, 31, 156–165.

[9] Trudinger, B. J., Giles, W. B., Cook, C. M., Bombardieri,J., Collins, L. E. E., Brit. J. Obstet. Gynecol. 1985, 92,23–30.

[10] Barker, D. J. P., Osmond, C., Winter, P. D., Margetts, B.,Simmonds, S. J., The Lancet 1989, 334, 577–580.

[11] Hales, C. N., Barker, D. J., Clark, P. M., Cox, L. J., Fall,C., Osmond, C., Winter, P. D., Brit. Med. J. 1991, 303,1019–1022.

[12] Phipps, K., Barker, D. J. P., Hales, C. N., Fall, C. H. D.,Osmond, C., Clark, P. M. S., Diabetologia 1993, 36, 225–228.

[13] Cecconi, D., Lonardoni, F., Favretto, D., Cosmi, E., Tucci,M., Visentin, S., Cecchetto, G., Fais, P., Viel, G., Ferrara,S. D., Electrophoresis 2011, 32, 3630–3637.

[14] Karamessinis, P. M., Malamitsi-Puchner, A., Boutsikou,T., Makridakis, M., Vougas, K., Fountoulakis, M., Vlahou,A., Chrousos, G., Mol. Cell. Proteom. 2008, 7, 591–599.

[15] Koy, C., Heitner, J. C., Woisch, R., Kreutzer, M., Serrano-Fernandez, P., Gohlke, R., Reimer, T., Glocker, M. O., Pro-teomics 2005, 5, 3079–3087.

[16] Heitner, J. C., Koy, C., Reimer, T., Kreutzer, M., Gerber,B., Glocker, M. O., J. Chromatogr. B 2006, 840, 10–19.

[17] Pecks, U., Seidenspinner, F., Rower, C., Reimer, T., Rath,W., Glocker, M. O., J. Am. Soc. Mass Spectrom. 2010, 21,1699–1711.

[18] Pecks, U., Schutt, A., Rower, C., Reimer, T., Schmidt, M.,Preschany, S., Stepan, H., Rath, W., Glocker, M. O., Hy-pertens. Pregnancy 2012, 31, in press.

[19] Voigt, M., Schneider, K. T., Jahrig, K., Geburtshilfe undFrauenheilkunde 1996, 56, 550–558.

[20] Hadlock, F. P., Harrist, R. B., Sharman, R. S., Deter, R. L.,Park, S. K., Am. J. Obstet. Gynecol. 1985, 151, 333–337.

[21] National High Blood Pressure Education Program Work-ing Group on High Blood Pressure in Pregnancy, Am. J.Obstet. Gynecol. 2000, 183, S1–S22.

[22] Wilcoxon, F., Biometrics Bull. 1945, 1, 80–83.

[23] Kruskal, W. H., Wallis, W. A., J. Am. Statist. Assoc. 1952,47, 583–621.

[24] Chambers, J. M., Hastie, T. J., Statistical Models inS, Wadsworth & Brooks/Cole, Pacific Grove, California1992.

[25] Stephens, M. A., J. Am. Statist. Assoc. 1974, 69, 730–737.

[26] Walpole, R. E., Myers, R., Probability & Statistics for En-gineers & Scientists, Macmillan Pubs, Hunts, UK, 1993.

[27] Tukey, J. W., Exploratory Data Analysis, Addison WesleyPub W Inc, Boston, MA, 1977.

[28] Kienbaum, M., Koy, C., Montgomery, H. V., Drynda, S.,Lorenz, P., Illges, H., Tanaka, K., Kekow, J., Guthke, R.,Thiesen, H.-J., Glocker, M. O., Proteomics – Clin. Appl.2009, 3, 797–809.

[29] Laemmli, U. K., Nature 1970, 227, 680–685.

[30] Bantscheff, M., Ringel, B., Madi, A., Schnabel, R.,Glocker, M. O., Thiesen, H.-J., Proteomics 2004, 4, 2283–2295.

[31] Eisen, M. B., Spellmann, P. T., Brown, P. O., Botstein, D.,Proc. Natl Acad. Sci. USA 1998, 95, 14863–14868.

[32] Schoonjans, F., Zalata, A., Depuydt, C. E., Comhaire, F.H., Computer Meth. Progr. Biomed. 1995, 48, 257–262.

[33] Koy, C., Mikkat, S., Raptakis, E., Sutton, C., Resch, M.,Tanaka, K., Glocker, M. O., Proteomics 2003, 3, 851–858.

[34] Koy, C., Resch, M., Tanaka, K., Glocker, M. O., Eur. J.Mass Spectrom. 2004, 10, 393–399.

[35] Bondarenko, P. V., Farwig, Z. N., McNeal, C. J.,Macfarlane, R. D., Int. J. Mass Spectrom. 2002, 219, 671–680.

[36] Ito, Y., Breslow, J. L., Chait, B. T., J. Lipid Res. 1989, 30,1781–1787.

[37] McComb, M. E., Oleschuk, R. D., Chow, A., Ens, W.,Standing, K. G., Perreault, H. l. N., Smith, M., Anal. Chem.1998, 70, 5142–5149.

[38] Nelsestuen, G. L., Harvey, S. B., Zhang, Y., Kasthuri,R. S., Sinaiko, A. R., Ely, E. W., Bernard, G. R., Homoncik,M., Jilma, B., Proteomics – Clin. Appl. 2008, 2, 158–166.

[39] Thompson, D., Develter, W., Cairns, D. A., Barrett, J. H.,Perkins, D. A., Stanley, A. J., Mooney, A., Selby, P. J.,Banks, R. E., Proteomics – Clin. Appl. 2011, 5, 523–531.

[40] Zurbriggen, K., Schmugge, M., Schmid, M., Durka, S.,Kleinert, P., Kuster, T., Heizmann, C. W., Troxler, H., Clin.Chem. 2005, 51, 989–996.

C© 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com

Page 13: Proteome Signature of IUGR From Cord Blood Serum

Electrophoresis 2012, 33, 1881–1893 Proteomics and 2-DE 1893

[41] Kuzyk, M. A., Smith, D., Yang, J., Cross, T. J., Jackson,A. M., Hardie, D. B., Anderson, N. L., Borchers, C. H.,Mol. Cell. Proteom. 2009, 8, 1860–1877.

[42] Pecks, U., Brieger, M., Schiessl, B., Bauerschlag, D. O.,Piroth, D., Bruno, B., Fitzner, C. T. O., Maass, N., Rath,W., J. Perinat. Med. 2012, 40, 287–296.

[43] Karlsson, H., Leandersson, P., Tagesson, C., Lindahl, M.,Proteomics 2005, 5, 551–565.

[44] Karlsson, H., Leandersson, P., Tagesson, C., Lindahl, M.,Proteomics 2005, 5, 1431–1445.

[45] Fredenrich, A., Diabetes Metab. 1998, 24, 490–495.

[46] Mann, C. J., Troussard, A. A., Yen, F. T., Hannouche, N.,Najib, J., Fruchart, J.-C., Lotteau, V., Andre, P., Bihain,B. E., J. Biol. Chem. 1997, 272, 31348–31354.

[47] Maeda, N., Li, H., Lee, D., Oliver, P., Quarfordt,S. H., Osada, J., J. Biol. Chem. 1994, 269, 23610–23616.

[48] Ito, Y., Azrolan, N., O’Connell, A., Walsh, A., Breslow,J. L., Science 1990, 249, 790–793.

[49] Tertov, V. V., Atherosclerosis 2001, 159, 103–115.

[50] Madi, A., Majai, G., Koy, C., Vamosi, G., Szanto, A.,Glocker, M. O., Fesus, L., Imumunol. Lett. 2011, 135, 88–95.

[51] Varki, A., Trends Mol. Med. 2008, 14, 351–360.

[52] Wang, B., Annu. Rev. Nutr. 2009, 29, 177–222.

[53] Schwarzkopf, M., Proc. Natl Acad. Sci. USA 2002, 99,5267–5270.

C© 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com


Recommended