an investigation of plasma collection, stabilization, and ...1:1:017.pdf · an investigation of...

Serum/Plasma Proteome

An Investigation of Plasma Collection,Stabilization, and Storage Procedures for Proteomic Analysis of Clinical SamplesJeffrey D. Hulmes,*,1 Deidra Bethea,1 Keith Ho,1 Shu-Pang Huang,2

Deborah L. Ricci,3 Gregory J. Opiteck1 Stanley A. Hefta1

1Department of Clinical Discovery, Proteomics Unit, Pharmaceutical Research Institute,Bristol-Myers Squibb Company, Princeton, NJ2,*Department of Clinical Discovery, Biostatistics Unit, Pharmaceutical Research Institute,Bristol-Myers Squibb Company, Princeton, NJ3,*Department of Clinical Discovery, Pharmacology Unit, Pharmaceutical Research Institute,Bristol-Myers Squibb Company, Princeton, NJ

17

Clinical Proteomics JournalCopyright ©Humana Press Inc.All rights of any nature whatsoever are reserved.ISSN 1542-6416/04/01:17–31/$25.00

*Author to whom all correspondence and reprint requests should be addressed: E-mail: [email protected]

AbstractIn order to evaluate the critical components

of the process necessary to preserve clinicalplasma samples collected at research sites forproteomic analysis, various collection and pre-servation protocols with controlled experi-mentation were evaluated. The presence of aprotease inhibitor cocktail (PIC) included in theblood draw tube would stabilize the plasmaproteins was hypothesized. To test this hypoth-esis, four plasma samples from each of 14 vol-unteers were collected. Samples were treatedfollowing a standard protocol that included PICor were subjected to various processing treat-ments that included time, temperature, differentanticoagulants, and the absence of PIC. Largeformat two dimensional-polyacrylamide gel

electrophoresis (2D-PAGE) proteomic analysisand enzyme immunoassay (EIA) were used todetect differences between the treatment groups.A novel 2D-PAGE quality scoring method wasdeveloped to determine global differences in thetreatment groups, wherein a rating scale ques-tionnaire was used to quantify the quality ofeach 2D-PAGE gel. The data generated fromEIAs, classical 2D-PAGE image analysis and 2D-PAGE quality scoring, each generated similarresults. Inclusion of protease inhibitor cocktailin the sample tubes, provided stable and reliablehuman plasma samples that yielded repro-ducible results by proteomic analysis. When PICwas included, samples retained stability underless stringent processing, such that refrigerationfor several hours before processing or onefreeze–thaw cycle had little detrimental effect.

Clinical Proteomics _______________________________________________________________ Volume 1, 2004

Key Words: Proteomics; plasma collection; two-dimensional gel electrophoresis; sample stability;protease inhibitors; 2D-PAGE evaluation.

Introduction

Proteomic analysis is now being used exten-sively for the study of disease through the useof clinical samples such as tumor cell extracts,tissue samples, or body fluids (1). Many ofthese studies involve the search for biomark-ers in which certain proteins change in expres-sion with either a large increase or decrease asa result of disease (1–4). Two-dimensionalgel electrophoresis (2-DE) remains the mostwidely used proteomic technique owing to thefact that it is currently the method that canseparate and measure hundreds of proteinsfrom a single biological sample simultane-ously (5,6). Analysis by 2D-PAGE has beenused for proteomic studies of breast cancer(7–9), prostate cancer (10,11), hepatocellularcarcinoma (12,13), renal cell carcinoma (14),rheumatoid arthritis (15,16), and hepatitis Binfection (17). Many of these biomarker stud-ies utilized human blood serum or plasmasamples. The plasma proteome is now beingwidely studied and will be a rich source ofproteins for proteomic analysis for years tocome (18).

The task of collecting large numbers of clini-cal plasma samples that best represent, andmost accurately reflect, the plasma compositionin vivo is of great importance. Given the largenumber of clinical trial sites around the world,it is important to standardize a procedure thatwill yield reliable samples for proteomic analy-sis. Clinical samples can be collected and pro-

cessed in any number of ways, but can also be“mishandled.” Plasma is often collected in avacuum drawn blood tube that contains anti-coagulants and a gel-separating barrier (e.g.VACUTAINER PPT™, Becton Dickinson,Franklin Lakes, NJ). There is a great possibilityof lab-to-lab variations in sample handling andprocessing and it would be easy to identify, forinstance, biomarkers of proteolysis rather thanmarkers of a particular disease. The stability ofsome analytes, viral load, and a few proteinsfrom collected plasma samples have been stud-ied (19–22), but to our knowledge the conditionor stability of the whole plasma proteome fromclinical samples has not been methodicallyexamined. To minimize the possibility of gen-erating “artifact” biomarkers, we have devel-oped a very precise protocol for the collectionof plasma samples from clinical trials. Thisprotocol includes the addition of a proteaseinhibitor cocktail (PIC) directly to plasma col-lection tubes prior to phlebotomy. Marshall etal. have recently shown that phenylmethyl-sulfonyl fluoride (PMSF) and ethylenediamine-tetraacetic acid (EDTA) effects the stability ofserum samples as judged by matrix assisitedlaser desorption/ionization mass spectrometry(MALDI-MS) (23) while Olivieri et al. have pre-viously demonstrated protease inhibitors in redblood cell membrane lysates can have largeeffect on 2D-PAGE analysis (24). Here we pre-sent our findings of controlled treatments ofplasma measured by proteomic analysis.

While it is logical that both a rigorous col-lection protocol and inclusion of PIC were nec-essary to obtain high quality plasma samplesfor proteomics analysis, neither had been thor-

18 _______________________________________________________________________________ Hulmes et al.

We demonstrated that samples without PIC, fromeither heparin or ethylenediaminetetraacetic acid(EDTA) plasma tubes, gave results that variedsignificantly from the control samples. Also, evenwith PIC present in blood tubes, we found it was

important to quickly decant the separated plasmafrom the cellular components found in the bloodtubes following centrifugation, as prolongedexposure again yielded different results from thestandard procedure.

Volume 1, 2004 _______________________________________________________________ Clinical Proteomics

oughly evaluated for their ability to preserveproteins. We therefore designed a series ofhandling treatments, with conditions onemight encounter in a clinical setting, to exam-ine their effects on protein stability and todetect differences from our standard protocol(Fig. 1). We wanted to verify that our methodgenerated stable and high quality plasma sam-ples. The variables studied centered aroundthe time and temperature of sample process-ing, the types of tubes used for sample collec-tion, and the presence or absence of PIC.

Proteomics is generally considered to be aglobal analysis of all proteins in a sample. Inthis study, we wanted to achieve the widestview of the plasma proteins to determine ifdifferences were present as a result of alter-ations in the standard protocol. We chose touse 2D-PAGE on the plasma samples in thisstudy. In order to limit the variables to mainly

the collection and storage differences, plasmasamples were run with a minimum of samplepreparation. Therefore, samples were notdepleted of high abundance proteins (i.e.,albumin and immunoglobulin) (25–27) , pre-fractionated, or run with narrow pI IPG strips(28–30). The goal was to run more that 50samples by duplicate 2D-PAGE, measure amaximum number of spots for each gel andcompare the control group to each of the othergroups treated by the various collection andprocessing parameters. For this reason wechose silver staining of the gels rather thanfluorescent staining, as the latter producedfewer overall spots, thus fewer data points toevaluate. Subsequently, the gels were analyzedusing classical image analysis (31,32) , as wellas a novel approach which we term Statistical2D-PAGE Quality Scoring (SQS). The SQSmethod was a semi-quantitative means of

Plasma Stabilization _________________________________________________________________________ 19

Fig. 1. Plasma samples collected and treated for study. A total of 56 samples were collected, 4 from eachvolunteer, and samples were treated by 8 different handling variations as described (A–H). All the samplescollected were analyzed by EIA and 2D-PAGE (except as noted in results).


objectively differentiating between good andbad quality gels. Different aspects and regionsof the gels were scored. Finally, immunoassaysfor single proteins, were used as a verificationmethod to test sample quality.

Materials and Methods

Experimental Design/Plasma CollectionProtocols

Four blood samples were drawn into tubesfrom 14 healthy volunteers for this studyyielding a total of 56 separate blood samplesfor proteomic analysis (Fig. 1). One sample ofblood from each patient was collected in atube that was pre-loaded with PIC and han-dled in accordance with our standard collec-tion and processing protocol (control groupA). For the control samples, plasma was col-lected in 8.0 mL lithium heparin plasma sepa-rating tubes (PST Vacutainer 367965, BectonDickinson, Franklin Lakes, NJ) preloaded with300 µL of PIC (stock: 2.2 mL of water addedto P-2714: Sigma, St. Louis, MO). Each controltube therefore contained lithium heparin,EDTA, AEBSF, bestatin, E-64, leupeptin, andaprotinin. Following blood collection into atube the following processing steps were con-ducted: (1) centrifuge at 2500g for 15 min at 4°Cwithin 15 min of the draw, (2) aliquot theplasma layer within 30 min of centrifugation in1.0 mL volumes, (3) freeze the aliquots imme-diately using a dry ice/alcohol bath, and (4)place frozen aliquots in –70°C freezer for stor-age. The goal was to have each plasma samplereach solid state within 1 h of venipuncture.

The remaining three blood samples fromeach volunteer were drawn and each sub-jected to alternate handling and storage treat-ments that would reasonably be encounteredat research sites participating in clinical trials.Each of the groups was a variation of the con-trol protocol and were labeled groups Bthrough H (six samples per group). For treat-

ment groups A through F, PIC preloadedVACUTAINERS™ were used, as described inthe previous section. In treatment group B, theblood was drawn and allowed to remain at4°C for 5 h prior to the initiation of centrifu-gation. The remaining steps were performedas described for the control. Treatment groupC varied from the control only by changingthe freezing temperature to –20°C. In treat-ment group D, the plasma was decanted aftercentrifugation, refrigerated as a bulk aliquot at4°C for 48 h before the generation of 1.0 mLaliquots subsequent freezing at –70°C. Treat-ment group E was processed by storing thecentrifuged VACUTAINER™ at 4°C for 48 hbefore aliquotting into 1.0 mL volumes andfreezing. Treatment group F was the result ofsnap freezing the decanted bulk plasma fol-lowing centrifugation, thawing after 48 h,aliquotting into 1.0 mL volumes, and re-freez-ing at –70°C. Treatment group G was collectedinto two EDTA tubes (VACUTAINER™, no.362788, 5.0 mL, Becton Dickinson) withoutaddition of PIC. Treatment group H excludedthe PIC from the lithium heparin VACUTAIN-ERS™ used in treatment group A, but other-wise remained unchanged.

RandomizationSamples were randomized prior to the 2D-

PAGE preparation step, the IEF procedure,and the second dimension run to allow formore robust statistical analysis. Multiple oper-ators were used for all steps of 2D-PAGEincluding staining and they were blinded tosample identification. Images for the SQSassessment were also randomized separatelyfor each scorer.

Sample Analysis

Two-Dimensional Electrophoresis and Gel Imaging

Sample preparation for 2D-PAGE was con-ducted according to a protocol described by

20 _______________________________________________________________________________ Hulmes et al.


the Two-Dimensional Gel ElectrophoresisLaboratory of Geneva, Switzerland (http://us.expasy.org/ch2d/service/) (33–35) with somemodification as detailed below. Each samplewas aliquotted in duplicate for two 2D-PAGEgels. Briefly, a 12.5 µL aliquot of plasma wasmixed with 10 µL of a solution containing 10%SDS (Fisher Scientific, Fair Lawn, NJ) and2.3% dithioerythritol (DTE) (Sigma, St. Louis,MO). The samples were heated to 95°C for5 min, and subsequently diluted to 500 µLwith solubilization buffer containing 8M urea(J.T. Baker, Phillipsburg, NJ), 4% Chaps (Cal-biochem, La Jolla, CA), 40 mM Tris (ICNBiomedicals, Aurora, OH), 65 mM DTE andtrace Bromphenol blue (Sigma). Protein loadswere based on the average plasma proteinconcentration of 80.0 ± 5.5 mg/mL from groupA samples as measured by bicinchoninic acid(BCA) protein assay (cat. no. 23227; Pierce,Rockford, IL) (data not shown).

Isoelectric focusing was performed on 18 cmpH 3.0–10.0 NL IPG strips (Amersham, Piscat-away, NJ), which were rehydrated overnightin buffer consisting of 8M urea, 2% Chaps 18mM DTE, 2% carrier ampholytes (Amersham)and bromphenol blue at room temperature.Sixty microliters of the prepared sample wasplaced into a cup loader at the anodic end ofthe IPG strip while at both electrodes, a paperwick immersed in a solution of 3.5% DTE wasplaced (26) . Each gel consisted of 120 µg ± 7%of plasma protein. Focusing was performed,12 strips at a time, on a Protean IEF Cell (Bio-Rad, Hercules, CA) for a total of 75kVh.

The second dimension sodium dodecyl sul-fate-polyacrylamide gel electrophoresis (SDS-PAGE) was performed on 20 × 25 cm 9–16%gradient Tris gels (Omnimatrix, Yardley, PA).The gels were run 12 at a time (Dodeca, Bio-Rad) at 100 V for 16.5 hr, followed by fixationin 40% methanol, 10% acetic acid for 1 h at25°C. The gels were silver stained by SilverExpress staining kit (cat. no. LC6100, Invitro-

gen, Carlsbad, CA) as described by the manu-facturer.

A cooled charge coupled device (CCD)camera (Fluor-S, cat. no. 1707704; Bio-Rad)was used to digitize the gel images, using thesame settings for all gels: f-stop = 5.6, expo-sure = 1.0 s, high resolution setting. Imageswere cropped to 757 by 978 pixels for unifor-mity and then were converted to TIFF files.These images were analyzed by ProgenesisDiscovery software (Version 2003.02; Nonlin-ear Dynamics, Durham, NC). The analysisexperiment was set up for automated spotdetection and matching, such that the gelsfrom each treatment set were groupedtogether and an average gel image file created.The average gel from control group A wasselected as the reference gel for the overallexperiment with all 108 gel images. Spotswere detected using the Progenesis algorithmand “Progenesis background” subtractionmodes. The combined warping and matchingsetting was used for gel matching, as was the“cross-gel detection” which “uses matchinginformation to make splitting more consistentwithin averaged gels.” Unmatched spots wereadded to the reference gel and spot numberssynchronized. The maximum “absence ofspots” allowed per treatment group was setto three. Normalized spot volumes wereobtained by normalizing to the total spotvolume (total density of image) after using theIntelligent Noise Correction Algorithm (INCA)for analysis. No attempt was made to manu-ally correct mismatched spots so as to keepthe objectivity of automated analysis.

Statistical 2D-PAGE Quality Scoring Analysis

The 108 randomized 2D-PAGE gel imageswere evaluated by five scoring analysts utiliz-ing a questionnaire with an equal-distancescoring scale (37) from poor (1) to fair (3) toexcellent (5). Scoring of gel images was doneat random with blinded scorers (2D-PAGE



operators) to limit the subjectivity (bias). Thescoring scale was used to judge nine criteria:(1) overall gel image, (2) overall focusing, (3)overall second dimension quality, (4) overallstreakiness, (5) general image contrast, (6)albumin/IgG resolution, (7) cluster chain res-olution, (8) low MW protein resolution, and(9) basic region resolution. The images wereeach scored against two reference gels,deemed to be of poor or excellent “quality.”Five scorers were asked to evaluate therandom-ordered gel images in sets of 20–22per scoring session. The scorers were blindedto operator and treatment information associ-ated with the images.

Protein ImmunoassaysEnzyme immunoassay (EIA) kits were

chosen based on commercial availability andapplicability to plasma samples. Immunoassaykits for P-Cadherin (PCAD, cat. no. DPCD0),and human soluble Vascular Cell AdhesionMolecule-1 (sVCAM-1, cat. no. BBE3), werepurchased from R & D Systems (Minneapolis,MN) and used according to the manufacturersinstructions. EIAs were performed in 96-wellplates for all 56 plasma samples in duplicate.Optical densities were determined by a platereader (cat. no. 550, Bio-Rad). Concentrationsof the proteins were calculated by executing amacro (XLFit, Excel, Microsoft, Redmond,WA) utilizing 4-parameter curve fits.

Statistical Analysis

Two-Dimensional Electrophoresis Scatter-Plot AnalysisThe spot volume data from matched aver-

age gels were exported from Progenesis toExcel and scatter-plots were created usingonly spots matched by automated matching.The data was plotted on a log–log scale andlinear regression analysis was done to producea linear equation as well as the correlationcoefficient (r2).

Statistical 2D-PAGE Quality Scoring Analysis

The scores from statistical 2D-PAGE qualityscoring analysis (SQS) of gel images werecompiled and analyzed by the cumula-tive logistic regression analysis (38) to obtainthe odds ratios for image differences by treat-ment effect. This gives a prediction of theprobability of one treatment group scoringhigher than another. All the odds ratios werecompared to treatment group A. The operatorand scorer effect were taken into considerationand were adjusted in the analysis.

Protein Immunoassay Analysis

The analysis of variance (ANOVA) wasapplied to test for differences among the treat-ment groups for each of the proteins measuredby immunoassay. Since four blood tubes weredrawn from each individual, the baseline levelfor proteins may be different between individ-uals. The baseline biological variation wasadjusted for evaluating the treatment effect inthe analysis to prevent confounding (39).

ResultsThe plasma samples for this study were col-

lected and processed according to the collec-tion protocol shown in Fig. 1. Each sampledrawn yielded at least one 1.0 mL aliquot ofplasma, with most yielding three aliquots.Immunoassays were performed on all 56 sam-ples and duplicate 2D-PAGE analysis wasdone on 54 samples for which there were ade-quate samples.

Following sample accrual and 2D-PAGEanalysis, all the gel images were processed ina single analysis set by the Progenesis soft-ware. An “average gel” was created from eachtreatment group, which calculated averagespot volumes for matched protein spots. Asummary of the 2D-PAGE image spot match-ing analysis where spots from control group Awere matched to each of the other treatmentgroups is shown in Table 1. For all 108 gels,

22 _______________________________________________________________________________ Hulmes et al.


between 318 and 835 spots were found foreach image. From the automated warping andmatching analysis, 174 to 200 spots from thecontrol group average gel were matched toeach of the other treatment group’s averagegel data. Figure 2 shows the scatter-plots foreach of the comparisons (control group A vsgroups B through H) using only the matchedspot average volumes. Variation betweengroups was determined by linear regressionanalysis and was used as a measure of samplestability. The coefficient of determination (r2

value) from scatter-plots of raw spot volumesand INCA normalized volumes are also plot-ted in Fig. 2. The results suggest that treat-ment groups G and H, which were collectedwithout PIC, are most different from the con-trol group. Groups E and F also show morescatter than the other groups when correlatedto the control group. As we expected, themore harshly the samples are treated, or in theabsence of PIC, the more scatter is seen in thespot volume comparisons.

We followed up the previous analysis witha second analysis of the gel images in order toget another perspective on the global differ-ences between treatment groups. A novelmethod was used to objectively score the qual-

ity of each gel image. Five analysts scorednine separate criteria, starting with overall gelimage impression and working towards selectportions of the gel (Fig. 3). The scores from theSQS process were compiled by treatmentgroup. Each treatment group was compared tocontrol group A. Significant odds ratios weredetected from the cumulative logistic regres-sion. Odds ratios and their 95% confidenceintervals for the “Overall Gel Image” are plot-ted in Fig. 4. For each scoring criteria, oddsratios were compiled and those with signifi-cant differences were ranked in order (Table2). The results indicated that treatments E, G,and H repeatedly showed significant differ-ences from treatment A (Table 2). Even thoughgroup E samples contained PIC, the gelimages from these samples ranked similarly totreatment group H samples that lacked PIC,thus leading us to conclude that it is veryimportant to decant the plasma quickly aftercentrifugation. Four of the treatment groupsstudied (B–D, and F) showed little differencefrom the control group (A). These groups allcontained PIC and were processed eitherwithin 6 h of the blood draw (treatments Band C) or the bulk plasma was removed fromthe collection tube (treatments D and F). In the


Table 12D-PAGE Image Analysis (Detection and Matching) for 108 Gels

Analyzed by Progenesis Discovery Software

Treatment Groups

A B C D E F G H

No. gels per group 126 112 112 110 112 112 112 112No. spots per

gel (range) 318–779 450–810 428–728 584–692 474–774 449–711 552–835 566–738No. of spots

matched per set 238 468 396 521 496 425 513 460(Average gels)No. spots matched 186 172 200 191 180 178 174

to group A spots

Fig.

2.Sc

atte

r-pl

ot a

naly

sis

of m

atch

ed a

vera

ge s

pot

volu

me

data

for

eac

h tr

eatm

ent

grou

p co

mpa

red

to c

ontr

ol g

roup

A.E

ach

com

pari

son

was

plo

tted

on

a lo

g–lo

g sc

ale

to c

reat

e a

scat

ter-

plot

.The

line

ar r

egre

ssio

n an

alys

is a

nd c

oeffi

cien

t of

det

erm

inat

ion

(r2 )

for

each

com

par-

ison

is s

how

n.D

ata

from

INC

A n

orm

aliz

ed s

pot

volu

me

in a

dditi

on t

o ra

w s

pot

volu

me

(sho

wn)

wer

e pl

otte

d an

d th

e r2

valu

es fr

om b

oth

sets

of

scat

ter-

plot

s ar

e gr

aphe

d in

the

low

er r

ight

pan

el.

24

Fig. 3. 2-D Gel Quality Scoresheet used for Statistical Quality Scoring (SQS). Each of 108 gels were randomlyscored by five scorers using the instructions and scoresheet shown here. Two reference gel images, one ofpoor and one of excellent quality, were chosen for the ends of the scoring scale by an experienced operatorwho did not take part in the scoring.


latter two conditions the bulk plasma, nowcontaining PIC, appeared to show no differ-ences even though samples sat at 4°C for 48 h(treatment D) or were exposed to afreeze–thaw cycle (treatment F). These datasuggest that plasma collected with PIC canwithstand temporary storage at 4°C, tempo-rary storage of plasma at –20°C, or at least asingle freeze–thaw cycle.

A series of immunoassays were performedon all 56 plasma samples to provide additionaldata to either support or refute our hypothe-sis. The assays were chosen because kits werecommercially available and were applicable toplasma samples. Assays for Placental Cad-herin (PCAD) and soluble Vascular AdhesionMolecule-1 (sVCAM-1) detected differencesbetween treatment groups when all sampleswere tested. The CVs for test plasma samplescollected using our standard protocol yielded

CVs for intra-assay and inter-assay precisionthat were equivalent to those described by themanufacturer (1.4–1.9% and 4.9–5.9% respec-tively for PCAD and 4.3–5.9% and 8.5–10.2%respectively for sVCAM-1 [data not shown]).To limit variability, samples were thawedonce, randomized and assayed in duplicate.The PCAD and sVCAM-1 immunoassayresults with descriptive statistics (mean andstandard error) for each treatment group areshown in Table 3. An analysis of variance(ANOVA) between control group A and eachof the seven other treatment groups (B–H)were applied to the immunoassay results(Table 4). After adjusting for baseline varia-tion, significant differences among treatmentswere observed. As a result, both PCAD andsVCAM-1 were shown to be present at signif-icantly higher levels in the plasma in treat-ment groups G and H (no PIC) as comparedwith groups that were drawn with PIC pre-sent. These two proteins have been reportedas potential biomarkers; PCAD has beenimplicated in invasive glandular lesions of thecervix and colon (40,41), whereas sVCAM-1has been linked to the development of coro-nary artery disease and atherosclerosis (42,43).

DiscussionThe overall goal of this work was to estab-

lish that our existing sample handling proto-col, including the protease inhibitor cocktail,stabilized the proteins in plasma samples col-lected for proteomic analysis. To test thishypothesis, three different experimental anal-yses were used on a total of 56 human plasmasamples collected from 14 separate donors.The samples were treated according to condi-tions specified in Fig. 1 in order to mimic con-ditions anticipated at clinical research sites.The analyses consisted of a classical spot anal-ysis of large format 2D gels, a novel scoringparadigm (SQS) for the same 2D gels, andimmunoassays for individual proteins.

26 _______________________________________________________________________________ Hulmes et al.

Fig. 4. The SQS scores for “Overall Gel Image” werecompiled, analyzed by cumulative logistic regression andthe odds ratios for each treatment group compared tocontrol group A are plotted with a 95% confidenceinterval (CI).Treatment groups E, G, and H were foundto be significantly different from the control group Aand are ranked according to odds ratio in Table 2.


The challenges of 2D-PAGE and image anal-ysis for clinical samples are many. 2D-PAGE isnot a high-throughput method, and imageanalysis—spot detection and matching—remains less than perfect. Also, because of thewide dynamic range and presence of highlyabundant proteins, plasma proteomics is amore difficult task than 2D-PAGE of bacterialextracts or even human cell extracts (18,28). Theclassical scatter-plot analyses, that were used tocompare each treatment group to the controlgroup (Fig. 2), represent a fairly crude analysistechnique. The statistical certainty was less thanideal, though there was a definite trend forsamples without PIC (G and H) to be differentfrom the control. Treatments E and F, where thesamples were treated more harshly, and possi-bly even B showed slightly more differencethan the groups C and D. While this result withimage analysis supported our hypothesis, rely-ing on the fit of a straight line to the matchedspots seemed to be an inaccurate means ofassessment. It was reassuring that groups G(EDTA only), and H (heparin only) showed themost differences from the control group, but itwas still less than statistically rigorous.

Therefore, an alternative method for statis-tically assessing gel separation quality wasdeveloped, so as to find meaningful differ-ences among treatment sets. The idea was touse the expertise of experienced 2D-PAGEoperators to grade the quality of the 2D-PAGEseparations, and therefore of the plasma sam-ples themselves. The result, that of the score-sheet and randomization process describedearlier (Fig. 3), were both surprisingly straight-forward and reassuringly simple. Unlike theresults from automated spot matching andscatter-plot analysis, the differences scoredbetween the control group (A) and treatmentgroups G and H were statistically significant(Table 2). The significant difference betweentreatment A and E (48 h refrigeration in col-lection tube after centrifugation), indicates thatthe plasma proteome was affected throughprolonged contact with the cellular material.One theory is that proteolytic enzymesexcreted from leukocytes diffused through thegel plug within the blood collection tube andoverwhelmed the PIC, causing proteolysis ofthe plasma proteins. Therefore, our collectionand processing protocols now ensure that the


Table 2Ranking of Significant Differences for Each of the Nine SQS Criteriaa

Criteria B vs A C vs A D vs A E vs A F vs A G vs A H vs A

Overall Gel Image 2 1 3Overall focusing 3 1 2Overall Second Dimension 1Overall Streakiness 3 1 2General Contrast 2 1 3Albumin/IgG Resolution 2 1Cluster Chain Resolution 3 2 1Low MW Protein Resolution 1 2 3Basic Region resolution 2 1

aOdds ratios with 95% confidence interval, like that plotted in Fig. 4, yielded significant differences which areranked for each scoring criteria. Rank is solely based on the odds ratio estimate, not statistically tested.

Note: Groups E, G, and H consistently scored significantly different than A in most of the nine quality crite-ria. There were no significant differences between groups B, C, or D compared to control group A.


sample tubes are centrifuged and decantedwithin 1 h of blood draw, and that this length oftime must be closely monitored and recorded.This data implicates that biomarker “artifacts”may result from extended incubation of bloodcomponents without separation. The results oftreatment groups G and H showed a benefitfrom the inclusion of the protease inhibitorsother than just EDTA. A possible future experi-ment would be to test blood samples for pro-tease activity directly (for example, usingZymogram gels). Finally, the nonsignificanceamong groups A–D and possibly F, indicatesthat the presence of PIC allowed for less expe-ditious processing and/or less rigorous storageconditions.

Results from gels from samples without PIC(G and H), and the samples left in a VACU-TAINER™ for 48 h at 4°C (E), were clearly rec-ognized by the SQS analysis as being distinctfrom the other 2D-PAGE separations. To thisend, the SQS process demonstrated its useful-ness. This methodology of scoring gel qualitymay be used for future clinical studies per-formed with 2D-PAGE analysis. It may also bepossible to use SQS as a first pass test to deter-mine if there are statistical differences in twosets of gels from a proteomic project prior tothe use of computed image analysis. The

results of SQS also proved to be readily man-ageable by statistical data analysis. A weightedscore for each of the scoring criteria can beeasily implemented to refine the SQS processto address more interesting questions.

Recognizing that the 2D-PAGE was a globalapproach, we searched for individual proteindifferences by spot volume analysis thatwould match the trends seen in the globalanalysis, however there was too much varia-tion in the data. By using immunoassays wedid find two proteins that changed signifi-cantly by treatment. These assays proved to bea high-throughput means of testing all sam-

28 _______________________________________________________________________________ Hulmes et al.

Table 3Results of Immunoassays of Human Soluble Placental cCadherin (PCAD) and Human Soluble Vascular

Adhesion Molecule-1 (sVCAM-1) for 56 Plasma Samples Collected for the Stability Studya

Treatment Groups

Protein Statistics A B C D E F G H

PCAD mean 9.203 9.451 9.022 10.088 9.645 9.030 11.094 11.494SD 1.691 1.615 1.742 2.217 1.875 1.582 0.908 1.367

sVCAM mean 241.993 235.487 196.315 227.756 273.396 282.871 766.919 396.786SD 99.457 89.596 92.618 85.906 133.166 89.381 239.811 109.271

aResults for mean concentration (ng/mL) with standard deviations (SD) for each treatment group are shown.

Table 4ANOVA Analysis (p values) of EIA Concentration

Values From Each Treatment GroupCompared to the Control Group

Contrast PCAD sVCAM-1

B vs A <0.4720 <0.9812C vs A <0.4623 <0.7504D vs A <0.2450 <0.8137E vs A <0.5575 <0.6152F vs A <0.2770 <0.7109G vs A <0.0001 <0.0001H vs A <0.0001 <0.0003


ples in a much shorter timeframe, and atsignificantly lower cost, than the 2D-PAGEmethod. The two proteins, PCAD andsVCAM-1, were known to be found in solubleand cell-bound forms, and we theorized thatan increase in the soluble form in plasmawould be because of proteolysis or cell lysisduring collection and handling of the bloodsamples. The fact that both PCAD andsVCAM-1 were found at higher levels in treat-ment groups G and H vs the control group Aindicates that both of these proteins are mostlikely cleaved by nonmetalloproteases andthat the presence of the PIC substantiallyincreases the stability of their membrane-bound forms. As surrogates for the rest of theplasma proteome, we infer that other proteinsare certain to be similarly affected by theabsence of protease inhibitors. The immunoas-say results corroborated the various 2D-PAGEgel analysis methods, in showing that treat-ment groups G (EDTA only) and H (no pro-tease inhibitors), the two groups without PIC,were again significantly different from thecontrol group A. The increase in solublePCAD or sVCAM-1 in these two groups sug-gests that they could serve as reporter proteinsfor proper collection or stability of plasmasamples. They may be surrogate markers ofplasma sample reliability.

Conclusions

As an overall assessment, each of the tech-niques and analysis methods showed remark-ably consistent results, that our standardcollection protocol, including the proteaseinhibitor cocktail, provides for the most stablehuman plasma samples for proteomic analy-sis. There were indications that some of thestrict aliquotting and time restraints couldbe relaxed, given the presence of the PIC,although there was clearly a need to separate

the cellular component of blood from theplasma as soon as possible, via centrifugationand decanting. Freeze–thaw cycles and tem-porary refrigeration both seemed to have littledetrimental affect in the presence of PIC.

The goals of the experiments also drove thecreation of the novel statistical scoring method(SQS). This method could be used in thefuture for the development of enhanced 2D-PAGE methods, both by vendors and end-users in clinical studies. It provides a robustmeans of differentiating subtleties in 2D-PAGEperformance, and can be implemented with-out the need for extensive computational anal-ysis hardware or software. We recognize thatthere are many new mass spectrometry tech-nologies being used for proteomics analysisbut 2D-PAGE analysis is still a commonand important tool for plasma and serum pro-filing. Not surprisingly, immunoassays forspecific proteins are a simple, fast, and inex-pensive way to validate sample stability orbiomarker assessment, limited only by kitavailability. Finally, if one adheres to ourprotocol for clinical plasma collection thatincludes protease inhibitors, even with thewide range of collection variability in researchlaboratories, there appears to be little need toworry about the integrity of the samples.

AcknowledgmentsThe authors are grateful for the assistance

of Elizabeth B. Hierl, Susan Lubin, and LisuWang, all of BMS-Hamilton, without whomthis project would not have been possible. Weare also indebted to Estelle M. Fach, Leah AnnGarulacan, Ji Gao, and Qing Xiao of BMS-Hopewell for their help running and scoringthe 2D-PAGE gels. Thanks also to ThomasColarusso and Christina Pavlick, of Nonlinearand Bio-Rad, respectively, who aided us in thearea of 2D-PAGE image analysis. Dr. Kim E.Zerba helped us tremendously with the bio-



statistics. Finally, JDH thanks Patricia E. Hulmesfor her valuable advice on questionnairescoring.

References1. Hanash S. Disease proteomics. Nature 2003;

422:226–232.2. Hanash SM. Biomedical applications of two-

dimensional electrophoresis using immobilizedpH gradients: current status. Electrophoresis2000;21:1202–1209.

3. Chambers G, Lawrie L, Cash P, Murray GI.Proteomics: a new approach to the study ofdisease. J Pathol 2000;192:280–288.

4. Seliger B, Kellner R. Design of proteome-based studies in combination with serologyfor the identification of biomarkers and noveltargets. Proteomics 2002;2:1641–1651.

5. Rabilloud T. Two-dimensional gel electro-phoresis in proteomics: old, old fashioned, butit still climbs up the mountains. Proteomics2002;2:3–10.

6. Young DS, Tracy RP. Clinical applications oftwo-dimensional electrophoresis. J Chroma-togr A 1995; 698:163–179.

7. Le Naour F, Misek DE, Krause MC, et al. Pro-teomics-based identification of RS/DJ-1 as anovel circulating tumor antigen in breastcancer. Clin Cancer Res 2001;7:3328–3335.

8. Rui Z, Jian-Guo J, Yuan-Peng T, Hai P, Bing-Gen R. Use of serological proteomic methodsto find biomarkers associated with breastcancer. Proteomics 2003;3:433–439.

9. Zuo X, Hembach P, Echan L, Speicher DW.Enhanced analysis of human breast cancerproteomes using micro-scale solution isoelec-trofocusing combined with high resolution1-D and 2-D gels. J Chromatogr B AnalytTechnol Biomed Life Sci 2002;782:253–265.

10. Charrier JP, Tournel C, Michel S, et al.Differential diagnosis of prostate cancerand benign prostate hyperplasia using two-dimensional electrophoresis. Electrophoresis2001;22:1861–1866.

11. Charrier JP, Tournel C, Michel S, Dalbon P,Jolivet M. Two-dimensional electrophoresis ofprostate-specific antigen in sera of men withprostate cancer or benign prostate hyperpla-sia. Electrophoresis 1999;20:1075–1081.

12. Poon TC, Johnson PJ. Proteome analysis andits impact on the discovery of serological

tumor markers. Clin Chim Acta 2001;313:231–239.

13. Steel LF, Shumpert D, Trotter M, et al. A strat-egy for the comparative analysis of serumproteomes for the discovery of biomarkers forhepatocellular carcinoma. Proteomics 2003;3:601–609.

14. Kellner R, Lichtenfels R, Atkins D, et al. Tar-geting of tumor associated antigens in renalcell carcinoma using proteome-based analysisand their clinical significance. Proteomics2002;2:1743–1751.

15. Sinz A, Bantscheff M, Mikkat S, et al. Massspectrometric proteome analyses of synovialfluids and plasmas from patients sufferingfrom rheumatoid arthritis and comparison toreactive arthritis or osteoarthritis. Electro-phoresis 2002;23:3445–3456.

16. Smith MA, Bains SK, Betts JC, Choy EH, Zan-ders ED. Use of two-dimensional gel elec-trophoresis to measure changes in synovialfluid proteins from patients with rheumatoidarthritis treated with antibody to CD4. ClinDiagn Lab Immunol 2001;8:105–111.

17. He QY, Lau GK, Zhou Y, et al. Serum bio-markers of hepatitis B virus infected liverinflammation: A proteomic study. Proteomics2003;3:666–674.

18. Anderson NL, Anderson NG. The humanplasma proteome: history, character, and diag-nostic prospects. Mol Cell Proteomics 2003;2:50.

19. Boyanton BL, Jr., Blick KE. Stability studies oftwenty-four analytes in human plasma andserum. Clin Chem 2002;48:2242–2247.

20. Flower L, Ahuja RH, Humphries SE,Mohamed-Ali V. Effects of sample handlingon the stability of interleukin 6, tumour necro-sis factor-alpha and leptin. Cytokine 2000;12:1712–1716.

21. Lewis MR, Callas PW, Jenny NS, Tracy RP.Longitudinal stability of coagulation, fibrinol-ysis, and inflammation factors in storedplasma samples. Thromb Haemost 2001;86:1495–1500.

22. Holodniy M, Rainen L, Herman S, Yen-Lieberman B. Stability of plasma humanimmunodeficiency virus load in VACUTAINERPPT plasma preparation tubes during overnightshipment. J Clin Microbiol 2000;38:323–326.

23. Marshall J, Kupchak P, Zhu W, et al. Process-ing of Serum Proteins Underlies the Mass

30 _______________________________________________________________________________ Hulmes et al.


Spectral Fingerprinting of Myocardial Infarc-tion. J Proteome Res. 2003;2:361–372.

24. Olivieri E, Herbert B, Righetti PG. The effectof protease inhibitors on the two-dimensionalelectrophoresis pattern of red blood cell mem-branes. Electrophoresis 2001;22:560–565.

25. Lollo BA, Harvey S, Liao J, et al. Improvedtwo-dimensional gel electrophoresis represen-tation of serum proteins by using ProtoClear.Electrophoresis 1999;20:854–859.

26. Wang YY, Cheng P, Chan DW. A simple affin-ity spin tube filter method for removing high-abundant common proteins or enrichinglow-abundant biomarkers for serum pro-teomic analysis. Proteomics 2003;3:243–248.

27. Rothemund DL, Locke VL, Liew A, ThomasTM, Wasinger V, Rylatt DB. Depletion of thehighly abundant protein albumin from humanplasma using the Gradiflow. Proteomics 2003;3:279–287.

28. Corthals GL, Wasinger VC, Hochstrasser DF,Sanchez JC. The dynamic range of proteinexpression: a challenge for proteomic research.Electrophoresis 2000;21:1104–1115.

29. Herbert B, Righetti PG. A turning point inproteome analysis: sample prefractionationvia multicompartment electrolyzers with iso-electric membranes. Electrophoresis 2000;21:3639–3648.

30. Zuo X, Echan L, Hembach P, et al. Towardsglobal analysis of mammalian proteomesusing sample prefractionation prior to narrowpH range two-dimensional gels and usingone-dimensional gels for insoluble and largeproteins. Electrophoresis 2001;22:1603–1615.

31. Collins PJ, Juhl C, Lognonne JL. Image analy-sis of 2D gels: considerations and insights.Cell Mol Biol (Noisy-le-grand) 1994;40:77–83.

32. Nishihara JC, Champion KM. Quantitative eval-uation of proteins in one- and two-dimensional

polyacrylamide gels using a fluorescent stain.Electrophoresis 2002;23:2203–2215.

33. Hughes GJ, Frutiger S, Paquet N, et al. Plasmaprotein map: an update by microsequencing.Electrophoresis 1992; 13:707–714.

34. Sanchez JC, Appel RD, Golaz O, et al. InsideSWISS-2DPAGE database. Electrophoresis1995;16:1131–1151.

35. Hoogland C, Sanchez JC, Walther D, et al.Two-dimensional electrophoresis resourcesavailable from ExPASy. Electrophoresis 1999;20:3568–3571.

36. Hoving S, Gerrits B, Voshol H, Muller D,Roberts RC, van Oostrum J. Preparative two-dimensional gel electrophoresis at alkaline pHusing narrow range immobilized pH gradi-ents. Proteomics 2002;2:127–134.

37. Streiner DL. A checklist for evaluating the use-fulness of rating scales. Can J Psychiatry 1993;38:140–148.

38. Agresti A. Analysis of Ordinal CategoricalData. New York: John Wiley & Sons, Inc.,1984.

39. Rhodes AE, Lin E, Streiner DL. Confrontingthe confounders: the meaning, detection, andtreatment of confounders in research. Can JPsychiatry 1999;44:175–179.

40. Han AC, Edelson MI, Peralta Soler A, et al.Cadherin expression in glandular tumors ofthe cervix. Cancer 2000;89:2053–2058.

41. Muller N, Reinacher-Schick A, Baldus S,et al. Smad4 induces the tumor suppressorE-cadherin and P-cadherin in colon carcinomacells. Oncogene 2002;21:6049–6058.

42. Blankenberg S, Rupprecht HJ, Bickel C, et al.Circulating cell adhesion molecules and deathin patients with coronary artery disease. Cir-culation 2001;104:1336–1342.

43. Cybulsky MI, Iiyama K, Li H, et al. A majorrole for VCAM-1, but not ICAM-1, in earlyatherosclerosis. J Clin Invest 2001;107:1255–1262.


an investigation of plasma collection, stabilization, and ...1:1:017.pdf · an investigation of...

Documents