Journal of Immunological Methods 396 (2013) 23–32

Contents lists available at ScienceDirect

Journal of Immunological Methods

j ourna l homepage: www.e lsev ie r .com/ locate / j im

Research paper

A new method combining sequential immunoaffinity depletionand differential in gel electrophoresis to identify autoantibodiesas cancer biomarkers

Marie Grandjean a, Marc Dieu b, Martine Raes b, Olivier Feron a,⁎a UCLouvain, Institut de Recherche Expérimentale et Clinique (IREC), Pole of Pharmacology and Therapeutics (FATH), Brussels, Belgiumb University of Namur, Namur Research Institute for Life Sciences (NARILIS), Laboratory of Biochemistry and Cellular Biology (URBC), Namur, Belgium

a r t i c l e i n f o

Abbreviations: aAbs, autoantibodies; TAA, tumorSERPA, serological proteome analysis; SID-DIGE, sequdepletion-differential in gel electrophoresis.⁎ Corresponding author at: UCLouvain, IREC, Pole

Therapeutics, 52 Avenue E. Mounier, B1.53.09, B-12Tel.: +32 2 7645264; fax: +32 2 7645269.

E-mail address: olivier.feron@uclouvain.be (O. Fer

0022-1759/$ – see front matter © 2013 Elsevier B.V. Ahttp://dx.doi.org/10.1016/j.jim.2013.07.006

a b s t r a c t

Article history:Received 26 May 2013Received in revised form 25 July 2013Accepted 26 July 2013Available online 2 August 2013

Easily measurable biomarkers are urgently required to detect early stages of cancer progression.Autoantibodies (aAbs), as a component of the humoral immune response against tumor cells,have such potential of diagnosticmarkers since they are circulating and stable proteins, producedrapidly and easily amenable to in vitro dosage. The identification of aAbs is based on thecharacterization of tumor-associated antigens (TAA) against which they are directed.Here,we propose a newmethod for an unbiased identification of TAA and thereby of aAbs as cancerbiomarkers. This method that we called sequential immunoaffinity depletion-differential in gelelectrophoresis (SID-DIGE) is based on the immunodepletion of tumor cell lysates with IgG fromcontrol and tumor-bearing mice and direct matching of the flow throughs of these immunoaffinityseparations on the same 2D format. This strategy reduces the complexity of the samples to beanalyzed and maximizes the interest of assessing hundreds of proteins simultaneously. SID-DIGEhas also the potential, contrary to existing serological proteome analysis (SERPA) techniques, todetect immunogenic proteins with conformational epitopes, including those resulting frompost-translationalmodifications. Using amodel of human colorectal tumors inmice for the proof ofprinciple, we showed that SID-DIGE outperforms the conventional SERPA technique, with theidentification of 7 common TAA (validating our approach) and 18 additional aAbs proving thepotential of this newmethod. In particular, the identification of aAbs directed against key enzymessupporting glycolysis gives credential to the role of hypoxia as a major determinant of the tumorproteome and thus as a source of immunogenicity. Overall, the developed methodology allowedefficient screening of sera for the identification of aAbs as potential biomarkers.

© 2013 Elsevier B.V. All rights reserved.

Keywords:AutoantibodiesSerological proteomeHypoxiaGlycolysis

1. Introduction

The term proteome describes the entirety of proteinsexpressed by a genome, with all functional posttranslational

-associated antigens;ential immunoaffinity

of Pharmacology and00 Brussels, Belgium.

on).

ll rights reserved.

modifications (Hanash and Taguchi, 2010). The cell proteomeis thus a dynamic field that changes fromminute to minute inresponse to intra- and extracellular environmental signals.Because blood perfuses all tissues of the body, serum containsthousands of components, indicators of diverse physiologicaland pathological states at a specific time (Petricoin et al.,2006). In the context of cancer, blood appears as an easilyaccessible source for biomarkers signing the presence of thedisease, its progression or the response to treatments (Cho,2007; Hanash et al., 2008). These markers can be of differentnatures including nucleic acids, peptides, antibodies, metab-olites and even circulating tumor cells (Hanash et al., 2008).

24 M. Grandjean et al. / Journal of Immunological Methods 396 (2013) 23–32

A good serum biomarker for cancer diagnosis and possiblyprognosis should be present at an early stage of the diseasedevelopment, be absent or at a low concentration in theblood of healthy subjects or patients suffering from otherpathologies and be easily quantifiable (Kulasingam andDiamandis, 2008). Practically, such biomarkers are producedby cancer cells and/or by surrounding tissues, in response tothe increase in tumor burden or because of microenviron-ment alterations such as inflammation and angiogenesis(Bouzin and Feron, 2007) as well as physicochemicalpeculiarities of growing tumors including low pH andhypoxia (Frerart et al., 2008; Daneau et al., 2010). Theidentification of protein biomarkers reflecting these differentaspects of tumor development and identifiable in the serumof patients faces the major obstacle of the wide range ofblood protein concentrations that spread on more than 10orders of magnitude and the huge abundance of a few highmolecular weight proteins masking the less abundant ones(Hanash et al., 2008).

The reduction of sample complexity is thus an essentialstep in the serum proteome analysis. Fractionation strategiesas specific depletion of highly abundant proteins can becarried out (Apweiler et al., 2009), but the elimination ofproteins such as albumin or immunoglobulins (Ig) inducesthe loss of proteins specifically or unspecifically linked tothem. Instead of trying to get rid of these abundant proteins,another possibility is to focus on them. In the search of cancerbiomarkers, this is particularly relevant for Ig since autoan-tibodies (aAbs) are part of the humoral immune responseagainst cancer. The production of aAbs directed againsttumor-associated antigens (TAA) was demonstrated in the1960s (for a review, see Tan et al., 2009; Desmetz et al.,2011). aAbs can be produced against “self-proteins” that aremutated, truncated, misfolded, over-expressed or ectopicallyexpressed by cancer cells during tumor development (Tanand Zhang, 2008; Reuschenbach et al., 2009). Althoughtumors develop different mechanisms to escape humoraland cellular responses of the host immune system (Kim et al.,2006), circulating aAbs represent a signature of cancerprogression. As biomarkers, aAbs present several advantages:they are stable proteins, they can be isolated through binding tospecific scaffolds such as protein A or G and they are easilydetectable using secondary antibodies. In addition, studies havedocumented that patients may carry aAbs many years beforethe manifestation of clinical symptoms (Lleo et al., 2010) andthat their detection could improve existing screening practices(Solassol et al., 2010) and enhance prognostic outcomes (Mangeet al., 2012).

Different methods exist to identify aAbs (Gunawardanaand Diamandis, 2007; Desmetz et al., 2009). Serologicalanalysis of recombinant cDNA expression libraries (SEREX)and serological proteome analysis (SERPA) are based on theuse of patient sera to probe blotted phage expression librariesderived from tumor cells or tumor cell lysates blotted onto amembrane after two-dimensional gel separation, respective-ly. The improvement of the SEREX technique involves phagedisplay to express proteins as fusions of the virion capsidproteins. Still, SERPA in contrast to SEREX and phage displayenables the detection of proteins that have undergonepost-translational modifications and allows the assessmentof 1000 to 3000 proteins simultaneously on a single format

(Defresne et al., 2010). The three major drawbacks of SERPAare however (i) the inability to detect immunogenic proteinswith conformational epitopes due to the denaturing condi-tions required for separating proteins on 2D gels, (ii) thecomplexity of the tumor cell lysates used as baits to identifyaAbs and (iii) the difficulty in matching several membraneslabelled with serum of either cancer patient or healthycontrols.

Here, we report the use of a new method to identify aAbsas cancer biomarkers exploiting the advantages of SERPA butavoiding its major inconveniences. This method that wecalled sequential immunoaffinity depletion-differential in gelelectrophoresis (SID-DIGE) allows the formation of immunecomplexes under native conditions, reduces the complexityof the samples to be analyzed and exploits the DIGEtechnology to automate the identification of TAA. For theproof of principle, we used a human colorectal cancer cell lineexposed to hypoxia and we directly compared the SID-DIGEtechnique to the SERPA method to validate some TAA butalso to document the advantage of the former on the latter.

2. Materials and methods

2.1. Cells

Human HCT 116 colorectal carcinoma cells were grown inMcCoy's 5A medium (Invitrogen) supplemented with 10%fetal bovine serum and antibiotics. Before lysate collection,cells were exposed for 48 h to hypoxia (1% O2) in a RuskinnInvivo2 500 workstation. Cells were washed with 20 mMsodium phosphate-buffered saline (PBS) and DIGE labelling(DLA) lysis buffer (7 M urea, 2 M thiourea, 4% CHAPS and30 mM Tris, pH 8.5) was used to scrape the cells.

2.2. Mice

Male 7-week-old NMRI mice (nu/nu) were purchasedfrom Elevage Janvier. 2 × 106 HCT116 cells were s.c. injectedin mice. When the mean tumor diameter reaches 8 mm,blood was collected by intra-cardiac route, and serum wasisolated following centrifugation at room temperature. Eachprocedure was approved by the local authorities, accordingto national animal care regulations.

2.3. IgG purification

IgGs were purified on an Akta micro-liquid chromatogra-phy system (GE Healthcare) using a HiTrap Protein G HPcolumn (1 ml) and buffers from the AB buffer kit (GEHealthcare). Pooled mouse serum (2 ml) was diluted with2 ml of binding buffer (0.2 M sodium phosphate, pH 7.0) andinjected on a pre-equilibrated column at a flow rate of0.5 ml/min. The column was then washed with 5 CV (columnvolumes) of binding buffer to remove unbound proteins. IgGswere eluted using the elution buffer (1 M glycine–HCl,pH 2.7), recovered with a fraction collector (Frac-950, GEHealthcare) and neutralized with 1 M Tris–HCl, pH 9.0. Thecolumn was finally washed with 3 CV of neutral bindingbuffer.

25M. Grandjean et al. / Journal of Immunological Methods 396 (2013) 23–32

2.4. Sequential immunoaffinity depletion

Purified control mouse IgGs (400 μg) were incubatedovernight at 4 °C with 4 mg of lysates of hypoxic HCT116cells to form immune complexes (under constant rotation).The sample was then applied on a HiTrap Protein G HPcolumn (1 ml) on the Akta micro LC (GE Healthcare) at0.5 ml/min. The column was washed with 5 CV of washingbuffer and the flow through (FT), corresponding to depletedlysates, was recovered. The putative immune complexes (IC)retained on the column were then eluted with 5 CV of elutionbuffer at 0.5 ml/min, recovered in a 96-well plate and rapidlyneutralized by 3 CV of neutralizing buffer. The flow through(FT1) was again incubated with 400 μg of purified controlmouse IgGs, and the mixture was chromatographed in thesame conditions to further deplete the tumor cell lysates(FT2); this procedure was repeated one more time to obtainFT3. The third flow through was then incubated with purifiedIgG from HCT116 tumor-bearing mice in order to deplete thetumor cell lysates of tumor-associated antigens; this proce-dure was repeated three times, thus giving rise at the end toflow throughs FT4, FT5 and FT6. Of note, before each newchromatography, elution buffer (3 CV) and binding buffer(3 CV) were used to wash and re-equilibrate the protein Gcolumn, respectively. In some experiments, IgG, FT and/or ICwere separated by SDS–PAGE and labelled with fluorescentKrypton protein stain (Pierce).

2.5. 2DE and SERPA

Tumor cell lysates were precipitated with a 2D Clean-upkit (GE Healthcare) and resuspended in DLA buffer.Twenty-five micrograms of extract was minimally labelledwith 200 pmol of cyanine dyes Cy5 (GE Healthcare) on ice for30 min in the dark, according to the manufacturer's protocol.The labelling reaction was stopped by incubating the mixturewith 10 mM lysine (Sigma Aldrich) for 10 min. Non-labelledproteins were added to reach a total of 250 μg of proteins anddiluted in a loading buffer (4% 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate (CHAPS), 7 M urea,2 M thiourea, 30 mM Tris, 30 mM DTT, 1% IPG buffer 3-11[GE Healthcare]) on ice for 15 min in the dark. Samples wereloaded onto rehydrated first dimension strips (ImmobilineDryStrip, pH 3-11 NL, 18 cm, GE Healthcare) and separatedon 2D gels, using isoelectric focusing in the first dimension(300 V for 3 h, gradient steps of 1000 V for 8 h, 8000 V for3 h, and 8000 V for 45 min at 20 °C with a maximum currentsettings of 50 μA per strip). In the second dimension, SDS–polyacrylamide gel (10% acrylamide) electrophoresis (SDS–PAGE) was run overnight at 15 °C. The proteins were finallytransferred onto a low-fluorescence PVDF membrane(200 mA for 1 h 30 min). Membranes were incubated for2 h in 5% non-fat dry milk-containing Tris buffer saline with1% of Tween (TTBS) blocking buffer and then blottedovernight at room temperature with either control ortumor-bearing mice serum (dilution 1/100); a pool of seracollected from 6 different mice was used per condition.Membranes were then incubated with horseradishperoxidase-conjugated anti-mouse IgG antibody (1/5000,Jackson Immunoresearch) for 2 h. Immunodetection wasperformed using ECL Plus (GE Healthcare) followed by

scanning on a Typhoon FLA9500 (GE Healthcare) at the Cy5wavelength for the detection of proteins and Cy2 wavelengthfor the Ab detection.

2.6. SID-DIGE

Proteins collected after sequential depletion using controlmouse IgG (FT 3) and tumor-bearing mouse IgG (FT6) werelabelled with Cy3 and Cy5 and inversely according to themanufacturer's protocol. A mix of the two extracts was alsominimally labelled with Cy2 and used as an internal standardin each loaded sample. All samples were loaded simulta-neously on the same first dimension strip (n = 4) andseparated as previously explained. After electrophoresis, thegels were scanned simultaneously at the Cy2, Cy3 and Cy5corresponding wavelengths and analyzed using the DeCyderDifferential Analysis Software v7.0 (GE Healthcare).

2.7. In gel enzymatic digestion and mass spectrometryidentification

For the identification of proteins of interest (from theSID-DIGE or SERPA method), unlabelled proteins wereseparated by 2DE and stained with Krypton (Pierce) afterprotein fixation in a 50% water, 40% ethanol and 10% aceticacid solution. Spots were automatically picked from thepreparative gels using an Ettan Spot Picker (GE Healthcare).After rinsing, the 2D gel picked spots were dehydrated inacetonitrile at 37 °C. Digestion was performed overnight at37 °C with trypsin (12.5 ng/ml) in 100 mM ammoniumbicarbonate. The digests were separated by reverse-phaseliquid chromatography using a 75 μm × 150 mm reversephase Dionex column (Acclaim PepMap 100 C18) in anUltiMate 3000 liquid chromatography system (Bruker).Mobile phase A was 95% of 0.1% formic acid in water and 5%acetonitrile. Mobile phase B was 0.1% formic acid inacetonitrile. The digest (15 μl) was injected, and the organiccontent of the mobile phase was increased linearly from 5% Bto 40% in 25 min and from 40% B to 100% B in 5 min. Thecolumn effluent was connected to a nano-LC-ESI-MS/MSmaXis 4G UHR-TOF (Bruker). In survey scan, MS spectra wereacquired for 0.5 s in the m/z range between 50 and 2200. The3 most intense peptides ions 2+ or 3+ were sequenced. Thecollision-induced dissociation (CID) energy was automatical-ly set according to mass to charge (m/z) ratio and chargestate of the precursor ion. Dionex and MaXis systems werepiloted by Compass HyStar 3.2 (Bruker). Peak lists werecreated using DataAnalysis 4.0 (Bruker) and saved as XMLfile for use with ProteinScape 2.1 (Bruker) with Mascot 2.2 assearch engine (Matrix Science). Enzyme specificity was set totrypsin, and the maximum number of missed cleavages perpeptide was set at one. Carbamidomethylation was allowed asfixed modification, oxidation of methionine and Gln-pyro-Gluwere allowed as variable modification. Mass tolerance formonoisotopic peptide window was 5 ppm and MS/MS toler-ance window was set to 0.05 Da. The NCBI nonredundantprotein database was searched with mammals as taxonomy.Only significant hits, as defined by the Mascot probabilityanalysis (p b 0.001), were accepted.

IgG IgG‘Tumor mouse’ IgG – Depleted lysate ( ) ‘Tumor mouse’ – Total lysate ( , ) ‘Control mouse’ – Total lysate ( , )

50KDa

25KDa

50KDa

BA

C

IC

ED

IgG FT IC1 IC2 IC3 IC4 IC5

25KDa

Fig. 1. Development and validation of the double depletion step supporting the SID-DIGE strategy. (A) SDS–PAGE separation of a purified fraction of IgG, theputative immune complexes (IC) and corresponding flow through (FT) resulting from the incubation of purified IgG and tumor cell lysates. (B) SDS–PAGEseparation of putative IC resulting from the incubation of purified IgG from control mice first with tumor cell lysates (IC1) and then successive flow throughs(IC2-IC4); IC5 corresponds to the incubation of purified IgG from tumor-bearing mice and the last depleted flow through (i.e., after IC4 removal). Note that mostcell lysate proteins interacting with control IgG were eliminated after the first incubation but that large amounts of immunoreactive proteins were foundfollowing consecutive incubation with IgG from tumor-bearing mice. (C) Representative pictures of 2D gel separation of IC resulting from the incubationsbetween the indicated purified IgG and tumor cell lysates. Insets correspond to magnification of the indicated parts of the 2D gels: green insets emphasize thepresence of proteins from total cell lysates which are recognized by IgG from both tumor-bearing and control mice while red insets show proteins from depletedor total cell lysates which interact with IgG from tumor-bearing mice but not with IgG from control mice.

26 M. Grandjean et al. / Journal of Immunological Methods 396 (2013) 23–32

3. Results

3.1. Immune complexes are formed by the combination ofpurified IgG and tumor cells lysates

HiTrap protein G affinity columns were first used to purifyIgG from pooled control or tumor-bearing mice sera. PurifiedIgG were then incubated with lysates of HCT116 colorectalcancer cells to form immune complexes (IC) in vitro. Theputative ICs were isolated on a new HiTrap Protein G column;the flow through (FT) or depleted lysates were also collected.The distinct patterns of purified IgG, purified IC and FT werevalidated following gel electrophoresis (Fig. 1A).

Fig. 2. Overview of the experimental SID-DIGE strategy. (A) Workflow of the SID-DIGcontrol mouse IgG with tumor cell lysates and then with successive flow throughs (with the last FT of the above depletion (FT3) and then with successive flow through(B) Typical 2D-DIGE analysis revealing the proteins present only in FT3 (green) or inor limitedly in FT6 because of specific immunodepletion with IgG from tumor-beacircle) is shown in the lower left panel while the right panel represents the mappinpreparative gels, for MS/MS protein identification.

3.2. Control mouse IgG-depleted lysates contain proteinsselectively interacting with tumor-bearing mouse IgG

HCT116 cell lysates were incubated with control mouseIgG and applied on the immunoaffinity chromatographysystem described above to selectively remove proteinsrecognized by circulating IgG of control mice (see lane 1 inFig. 1B). Of note, in these experiments, lysates were derivedfrom HCT116 cells pre-incubated for 48 h under hypoxia tobetter mimic the in vivo tumor microenvironment andthereby favor the expression of proteins not or poorlyexpressed under usual (normoxic) culture conditions. Therecovered FT was then repeatedly exposed to new pools of

E depicting the double depletion resulting (i) from the incubation of purifiedFT1-3) and (ii) from the incubation of purified IgG from tumor-bearing mices (FT4-5). Aliquots of FT3 and FT6 are then used to perform 2D-DIGE analysisboth FT3 and FT6 (yellow); spots of interest are those present in FT3 but noring mice. An example of software-based analysis of a spot of interest (redg of the 37 spots displaying differential intensity, that were processed from

.t

27M. Grandjean et al. / Journal of Immunological Methods 396 (2013) 23–32

control IgG to optimize the depletion. As shown in Fig. 1B(compare lanes 2-4), most proteins interacting with controlIgG were eliminated during the first depletion step. The last

control mouse IgG

hypoxic tumlysate

mix

HiTrap Prot G

IC

FT1FT2FT3

tumor-bearing mouse IgG

mix

HiTrap Prot G

FT4FT5FT6

A

IC

B 3

1

28

FT3 FT6

FT was then incubated with circulating IgG collected fromtumor-bearing mice and the formation of immune complexeswas confirmed by gel electrophoresis (Fig. 1B, last lane).

or cells

2D-DIGEautomated comparison

pI 11

12

13

93 4

6

25

14

22

7

8

18

10

11

12

205

16

17

19

21

5

23

24

26

2729

3032

31

33

34

35

36

37

28 M. Grandjean et al. / Journal of Immunological Methods 396 (2013) 23–32

To further validate the depletion procedure, we comparedthe 2D-pattern of immune complexes formed following theincubation of depleted lysates with IgG from tumor-bearingmice (Fig. 1C), of total lysates with IgG from tumor-bearingmice (Fig. 1D) and of total lysates with control mouse IgG(Fig. 1E). Some proteins from total lysates are recognized byIgG from both tumor-bearing and control mice (comparegreen insets of Figs. 1D and 1E). More importantly, proteinsfrom the depleted lysates can interact with IgG fromtumor-bearing mice while these spots are not found whenIgG from control mice are incubated with total lysates(compare red insets of Figs. 1C and E).

3.3. SID-DIGE technique outperforms SERPA in the identificationof tumor-associated antigens

We then used 2D-DIGE technology to compare lysates ofhypoxic HCT116 cells depleted either with purified IgG fromcontrol mice or with purified IgG from both control and

Table 1Immunogenic proteins from HCT116 human colorectal tumors identified using the

Number Protein A

Glucose metabolism4 Glyceraldehyde-3-phosphate dehydrogenase P10 Transaldolase P11 Alpha-enolase P12 Voltage-dependent anion-selective channel protein 1 P13 Alpha-enolase P14 Alpha-enolase P16 Glyceraldehyde-3-phosphate dehydrogenase P17 Fumarate hydratase, mitochondrial P20 L-Lactate dehydrogenase A chain P25 Glyceraldehyde-3-phosphate dehydrogenase P29 Alpha-enolase P30 Pyruvate kinase P35 Alpha-enolase P36 Fructose-bisphosphate aldolase A P

Chaperones5 90-kDa heat shock protein P7 HSP90AA1 protein Q8 HSP90AA1 protein Q22 90-kDa heat shock protein P23 T-complex protein 1 subunit beta P26 DnaJ protein homolog P34 60-kDa heat shock protein, mitochondrial P

Cytoskeleton related1 Cytokeratin 18 P15 Tubulin beta chain P24 Tubulin alpha-1A chain Q28 Cytokeratin 8 Q31 Ezrin Q32 Ezrin Q

Annexins2 Annexin A1 P3 Annexin A2 P19 Annexin A2 P21 Annexin A2 P

Protein synthesis/degradation/translocation6 Eukaryotic translation elongation factor 2 Q9 60S acidic ribosomal protein P0 P18 Eukaryotic initiation factor 4A-I isoform 1 A27 Nucleolin P33 26S protease regulatory subunit 8 P37 GTP-binding nuclear protein Ran P

tumor-bearing mice. To do so, we compared the last flowthroughs of a series of 3 immunopurifications using controlIgG (FT3) and of an additional series of 3 immunopurifi-cations using tumor-bearingmouse IgG (FT6) (Fig. 2A). After2D-separation of differentially labelled proteins (seeMethods), gels were scanned and analyzed with the DeCydersoftware (Fig. 2B, left panels). Eighty spots displayed a≥2-folddecrease in abundance in the double-depleted lysates, i.e., FT6(vs. FT3), and 37 of them could be further picked on pre-parative gels (Fig. 2B, right panel). Mass spectrometry analysisled to the identification of 25 different proteins with asatisfying Mascot score (p b 0.001) (Table 1).

We then compared the SID-DIGE strategy (i.e., the2D-DIGE comparison of sequentially depleted lysates) andserological proteome analysis (SERPA). For that purpose,lysates from hypoxic HCT116 tumor cells were labelled withCy5 and separated through 2D electrophoresis beforetransfer on PVDF membranes (Fig. 3A). Pooled sera fromcontrol or tumor-bearing mice were then incubated with the

SID-DIGE technology and MS/MS analysis (p b 0.001).

ccession no. Mascot scores Peptides Fold change

04406 970 17 −3.4737837 68 2 −2.9206733 662 13 −2.8621796 172 4 −2.8606733 883 16 −2.7806733 1341 21 −2.6504406 970 17 −2.67954 300 5 −2.5400338 570 12 −2.4704406 915 17 −2.3606733 547 9 −2.2214618 2501 43 −2.2106733 806 13 −2.0404075 168 2 −2.04

08238 1030 19 −3.496HX7 677 12 −2.9896HX7 887 16 −2.9408238 811 13 −2.4278371 262 5 −2.4131689 1092 20 −2.3410809 286 6 −2.06

05783 911 19 −5.5807437 393 9 −2.6271U36 668 12 −2.36969I0 477 9 −2.236NUR7 378 10 −2.176NUR7 1689 32 −2.14

04083 944 15 −4.1707355 1023 22 −3.6807355 968 19 −2.4907355 134 2 −2.47

6PK56 72 2 −3.0805388 1039 17 −2.948K7F6 666 11 −2.5307954 494 12 −2.3162195 1085 18 −2.1262826 307 8 −2.01

Hypoxic tumor cell lysates

+ controlmouse IgG

2D-gel membrane

transfer

+ tumor-bearingmouse IgG

A

+ 2nd Ab

+ 2nd Ab

eyecomparison

B pI3 11

1

4

79

6

8

2

3

5

Hypoxic cell lysates –

control mouse serum

Hypoxic cell lysates –

tumor-bearing mouse serum

Fig. 3. Overview of the experimental SERPA strategy. (A) Workflow depicting the SERPA process of 2D gel separation of tumor cell lysates and after membranetransfer, the eye comparison of the immunoblotting resulting from the incubation with serum from either control or tumor-bearing mice. (B) Typical 2D gels withthe distinct patterns of immunoblotting corresponding to the incubation of hypoxic cell lysates with the indicated mouse serum (left panels) and mapping of the9 spots displaying differential intensity that were processed from preparative gels, for MS/MS protein identification (right panel).

29M. Grandjean et al. / Journal of Immunological Methods 396 (2013) 23–32

membranes (Fig. 3A) and antigenic recognition by the cor-responding sera was detected using a peroxidase-conjugatedsecondary antibody (Fig. 3B, left panels). The comparison ofthe immunoblotting signals in each condition led to theidentification of 9 spots specifically recognized by the sera oftumor-bearing mice (Fig. 3B and Table 2). Mass spectrometryanalysis led to the identification of these 9 proteins of interestwith a satisfying Mascot score (p b 0.001) (Table 2); 7 ofthem had also been identified by the SID-DIGE strategy (seebolded names in Tables 1 and 2).

4. Discussion

In this study, we developed a new workflow to detectimmunogenic TAA and thereby identify potential cancerbiomarkers. We used affinity chromatography to successivelydeplete samples of non-tumor-specific antigens and tumor-associated antigens, and performed a differential comparisonof the corresponding samples (i.e., flow throughs) devoid ofcontaminating antibodies. We named this strategy SID-DIGEto integrate the concepts of sequential immunoaffinitydepletion (SID) and 2D-differential in gel electrophoresis

(2D-DIGE). Contrary to the conventional SERPA and SEREXmethodologies aiming to identify aAbs (see Introduction), theSID-DIGE strategy offers the advantages to allow the forma-tion under native conditions, of immune complexes betweenaAbs and TAA, including potentially post-translationnallymodified proteins, and to exploit 2D-DIGE to automatedlyidentify candidate biomarkers by comparing decomplexifiedmixtures of TAA separated in a same gel.

The purification of circulating antibodies and immunecomplexes is based on the high affinity of protein G fromStreptococci for the Fc region of IgG, a procedure easilyautomated using microscale LC system in our study. Noelution method however allows to quantitatively and selec-tively release the antigens from the antibodies fixed on thecolumn. The elution methods based on the alteration of thepH or ionic strength of the buffer eventually lead to thecollection of both antigens and antibodies in the eluate. After2D gel separation, the large amounts of IgG reduce thedynamic range of antigen detection (see Figs. 1C–E whereonly a few highly expressed proteins are detectable).Although crosslinking may be used to irreversibly fix theantibodies on the column, the yield of this procedure is not

Table 2Immunogenic proteins from HCT116 human colorectal tumors identified using the SERPA technology and MS/MS analysis (p b 0.001).

Number Protein Accession no. Mascot score Peptides

Glucose metabolism5 Fumarate hydratase, mitochondrial P07954 856 116 Glyceraldehyde-3-phosphate dehydrogenase P04406 245 37 Transaldolase P37837 520 88 Fructose-bisphosphate aldolase A P04075 75 19 Voltage-dependent anion-selective channel protein 2 P45880 252 3Chaperones4 T-complex protein 1 subunit beta P78371 2318 30Stress response2 N-myc downstream-regulated gene 1 protein Q92597 1163 13Cytoskeleton3 Cytokeratine 8 P05787 1584 21Protein synthesis1 Eukaryotic translation elongation factor 2 P13639 1969 26

30 M. Grandjean et al. / Journal of Immunological Methods 396 (2013) 23–32

compatible with a quantitative method. In this study, wetherefore chose to compare the flow throughs of thesuccessive depletion steps using purified IgG from controlmouse and tumor-bearing mouse sera. This strategy allowedto get rid of the antibodies retained on the HiTrap Protein Gcolumns and to exploit 2D-DIGE to detect immunogenicproteins present in the flow through of the mixture of celllysates with control mouse IgG but which disappearedfollowing exposure to tumor-bearing mouse IgG.

The SID-DIGE strategy led to the identification of 7 of the 9immunogenic proteins found with the SERPA method andimportantly further pinpointed 18 additional TAA. Besidesthe antigen–antibody interaction occurring under nativeconditions (vs. the antibodies hybridization on denaturedand blotted antigens for the SERPA method), the SID-DIGEapproach takes advantage of the power of an automated,statistically meaningful method for identifying immunogenicproteins; with the SERPA approach, matching of membranesarising from distinct 2D gels needs indeed to be done by eye.Also, in our workflow, the elimination of IC resulting from thepre-incubation with control IgG reduces the complexity ofthe lysate proteome and thus increases the relative abun-dance of proteins of interest, thereby improving the detectionof less abundant antigens. Of note, for the current proof ofprinciple, we only picked up proteins in spots with adifferential intensity of N2 between FT3 and FT6, indicatingthat additional potential TAA could actually be identified.

Among the immunogenic proteins identified, 5 enzymesof the glycolytic pathwayswere identified:α-enolase, glyceral-dehyde-3-phosphate dehydrogenase, fructose-bisphophatealdolase, pyruvate kinase and lactate dehydrogenase A (Fig. 4).Some of these enzymes were already reported to be immuno-genic in a variety of cancers (Ludwig et al., 2012). Still, thedetection of five of them indicates that this metabolic pathwayis very critical for cell survival and proliferation under thehypoxic conditions imposed to colorectal cancer cells used inthis study. We used this hypoxic challenge to mimic the tumormicroenvironment known to be hypoxic in a large majority oftumors. Tumor cells adapt to low pO2 through a generaldecrease in energy-consuming processes such as transcriptionand translation and the induction of a specific gene expressionprogram (Feron, 2009). The latter is largely dependent on thefamily of transcription factors HIF (hypoxia-inducible factors)which promote angiogenesis (Vincent et al., 2002) but also

support a metabolic switch facilitating the production ofanabolic intermediates and ATP independently of oxidativephosphorylations (Feron, 2009; Polet and Feron, 2013). Theoverexpression of glycolytic enzymes to the detriment of manyother downregulated signalling proteins is therefore very likelyto account for the production of corresponding aAbs as detectedin our study. In addition, two other enzymes, transaldolaserequired for the non-oxidative component of the pentosephosphate pathway and fumarate hydratase involved in theTCA cycle, were also identified as the targets of aAbs in ourstudy (Fig. 4). While the former promotes the formation offructose-6-phosphate from 3 carbon sugars (and thereby canfeed the glycolytic pathway), the second leads to the productionof malate, which may give rise to NADPH following cytosolicconversion into pyruvate (thus without the need of coupling tothe respiratory chain). Finally, a last immunogenic proteinidentified in our study is the voltage-dependent anion-selectivechannel (VDAC) (Fig. 4), which is described as a pore requiredfor regulating the exchange of nucleotides and a variety ofmetabolites in and out of the mitochondria but also as afunctional anchor for the glycolytic enzyme hexokinase(Mathupala et al., 2006).

5. Conclusions

The SID-DIGE strategy offers several benefits over othertechnologies for aAbs identification, including non-denaturingconditions for the antigen–antibody interaction, enrichment inlow abundance immunogenic proteins, reduction of antigenicdiversity and elimination of contaminating antibodies in the 2Dgels, which improved data analysis using automation anddedicated softwares. In this study, the SID-DIGE strategy ledto the identification of enzymes supporting the glycolyticswitch associated with hypoxia as major immunogenic pro-teins leading to the production of corresponding aAbs signingthe tumor progression in mice.

Acknowledgements

This work was supported by grants from the FondsNational de la Recherche Scientifique (F.R.S.-FNRS), theFonds de la Recherche Scientifique Médicale (FRSM), anAction de Recherche Concertée from the Communautéfrançaise de Belgique (grant no. ARC 09/14-020), the Belgian

Pentose Phosphate Pathway

Ribulose-5-Phosphate

Ribose-5-Phosphate

Fumarate

Malate

Malate TCA cycle

Ox

Non-ox

Glucose

Glucose-6-P

Fructose-6-P

Fructose-1,6-P

Dihydroxyacetone-P

Glyceraldehyde-3-P

1,3-bisphosphoglycerate

2-phosphoglycerate

Phosphoenolpyruvate

Pyruvate

Lactate

ENO

LDHA

TALDO

GAPDH

VDAC

ALDO

PKM

HK

FH

Mitoch.

VDAC

Fig. 4. Mapping of the immunogenic proteins related to the glucose metabolism in hypoxic tumor cells. In red are shown the identified five enzymes of theglycolytic pathway, namely, fructose-bisphophate aldolase (ALDOA), glyceraldehyde-3-phosphate dehydrogenase (GAPDH), α-enolase (ENO), pyruvate kinase(PKM) and lactate dehydrogenase A (LDHA), as well as enzymes from the pentose phosphate pathway and TCA cycle, transaldolase (TALDO) and fumaratehydratase (FH), respectively, and the voltage-dependent anion-selective channel (VDAC), the mitochondrial porin, also described as a scaffold protein for theglycolytic enzyme hexokinase (HK).

31M. Grandjean et al. / Journal of Immunological Methods 396 (2013) 23–32

Science Policy Office (Interuniversity Attraction Pole (IUAP)grant no. UP7-03) and the J.Maisin Foundation.M.Grandjean isa FRIA (Fonds pour la formation à la Recherche dans l'Industrieet dans l'Agriculture) Research Fellow. The MS facility of theURBC-NARILIS is supported by the FNRS (Brussels, Belgium).

References

Apweiler, R., Aslanidis, C., Deufel, T., Gerstner, A., Hansen, J., Hochstrasser, D.,Kellner, R., Kubicek, M., Lottspeich, F., Maser, E., Mewes, H.W., Meyer,H.E., Mullner, S., Mutter, W., Neumaier, M., Nollau, P., Nothwang, H.G.,Ponten, F., Radbruch, A., Reinert, K., Rothe, G., Stockinger, H., Tarnok, A.,Taussig, M.J., Thiel, A., Thiery, J., Ueffing, M., Valet, G., Vandekerckhove, J.,Verhuven, W., Wagener, C., Wagner, O., Schmitz, G., 2009. Approachingclinical proteomics: current state and future fields of application in fluidproteomics. Clin. Chem. Lab. Med. 47, 724.

Bouzin, C., Feron, O., 2007. Targeting tumor stroma and exploiting maturetumor vasculature to improve anti-cancer drug delivery. Drug Resist.Updat. 10, 109.

Cho, W.C., 2007. Contribution of oncoproteomics to cancer biomarker dis-covery. Mol. Cancer 6, 25.

Daneau, G., Boidot, R., Martinive, P., Feron, O., 2010. Identification ofcyclooxygenase-2 as a major actor of the transcriptomic adaptation ofendothelial and tumor cells to cyclic hypoxia: effect on angiogenesis andmetastases. Clin. Cancer Res. 16, 410.

Defresne, F., Bouzin, C., Guilbaud, C., Dieu, M., Delaive, E., Michiels, C., Raes,M., Feron, O., 2010. Differential influence of anticancer treatments andangiogenesis on the seric titer of autoantibody used as tumor andmetastasis biomarker. Neoplasia 12, 562.

Desmetz, C., Maudelonde, T., Mange, A., Solassol, J., 2009. Identifying auto-antibody signatures in cancer: a promising challenge. Expert. Rev. Proteomics6, 377.

Desmetz, C., Mange, A., Maudelonde, T., Solassol, J., 2011. Autoantibodysignatures: progress and perspectives for early cancer detection. J. CellMol. Med. 15, 2013.

Feron, O., 2009. Pyruvate into lactate and back: from the Warburg effect tosymbiotic energy fuel exchange in cancer cells. Radiother. Oncol. 92, 329.

Frerart, F., Sonveaux, P., Rath, G., Smoos, A., Meqor, A., Charlier, N., Jordan, B.F.,Saliez, J., Noel, A., Dessy, C., Gallez, B., Feron, O., 2008. The acidic tumormicroenvironment promotes the reconversion of nitrite into nitric oxide:towards a new and safe radiosensitizing strategy. Clin. Cancer Res. 14, 2768.

Gunawardana, C.G., Diamandis, E.P., 2007. High throughput proteomic strategiesfor identifying tumour-associated antigens. Cancer Lett. 249, 110.

Hanash, S., Taguchi, A., 2010. The grand challenge to decipher the cancerproteome. Nat. Rev. Cancer 10, 652.

Hanash, S.M., Pitteri, S.J., Faca, V.M., 2008. Mining the plasma proteome forcancer biomarkers. Nature 452, 571.

Kim, R., Emi, M., Tanabe, K., Arihiro, K., 2006. Tumor-driven evolution ofimmunosuppressive networks during malignant progression. CancerRes. 66, 5527.

Kulasingam, V., Diamandis, E.P., 2008. Strategies for discovering novel cancerbiomarkers through utilization of emerging technologies. Nat. Clin.Pract. Oncol. 5, 588.

32 M. Grandjean et al. / Journal of Immunological Methods 396 (2013) 23–32

Lleo, A., Invernizzi, P., Gao, B., Podda,M., Gershwin,M.E., 2010. Definition of humanautoimmunity–autoantibodies versus autoimmune disease. Autoimmun. Rev.9, A259–A266.

Ludwig, N., Keller, A., Leidinger, P., Harz, C., Backes, C., Lenhof, H.P., Meese, E.,2012. Is there a general autoantibody signature for cancer? Eur. J. Cancer48, 2451.

Mange, A., Lacombe, J., Bascoul-Mollevi, C., Jarlier, M., Lamy, P.J., Rouanet, P.,Maudelonde, T., Solassol, J., 2012. Serum autoantibody signature ofductal carcinoma in situ progression to invasive breast cancer. Clin.Cancer Res. 18, 1992.

Mathupala, S.P., Ko, Y.H., Pedersen, P.L., 2006. Hexokinase II: cancer's double-edged sword acting as both facilitator and gatekeeper of malignancywhen bound to mitochondria. Oncogene 25, 4777.

Petricoin, E.F., Belluco, C., Araujo, R.P., Liotta, L.A., 2006. The blood peptidome: ahigher dimension of information content for cancer biomarker discovery.Nat. Rev. Cancer 6, 961.

Polet, F., Feron, O., 2013. Endothelial cell metabolism and tumour angiogenesis:glucose and glutamine as essential fuels and lactate as the driving force.J. Intern. Med. 273, 156.

Reuschenbach, M., von Knebel, D.M., Wentzensen, N., 2009. A systematicreview of humoral immune responses against tumor antigens. CancerImmunol. Immunother. 58, 1535.

Solassol, J., Rouanet, P., Lamy, P.J., Allal, C., Favre, G., Maudelonde, T., Mange,A., 2010. Serum protein signature may improve detection of ductalcarcinoma in situ of the breast. Oncogene 29, 550.

Tan, E.M., Zhang, J., 2008. Autoantibodies to tumor-associated antigens:reporters from the immune system. Immunol. Rev. 222, 328.

Tan, H.T., Low, J., Lim, S.G., Chung, M.C., 2009. Serum autoantibodies asbiomarkers for early cancer detection. FEBS J. 276, 6880.

Vincent, K.A., Feron, O., Kelly, R.A., 2002. Harnessing the response to tissuehypoxia: HIF-1 alpha and therapeutic angiogenesis. Trends Cardiovasc.Med. 12, 362.