A catalogue of proteins released by colorectal cancer cellsin vitro as an alternative source for biomarker discovery

RESEARCH ARTICLE

A catalogue of proteins released by colorectal

cancer cells in vitro as an alternative source for

biomarker discovery

Hanna C. Diehl1, 2*, Kai Stühler1*, Susanne Klein-Scory2, Martin W. Volmer2,Anna Schöneck2, Cornelia Bieling1, Wolff Schmiegel2, 3, Helmut E. Meyer1 andIrmgard Schwarte-Waldhoff2

1 Medical Proteome-Center, University of Bochum, Bochum, Germany2 Department of Internal Medicine, IMBL, Knappschaftskrankenhaus, University of Bochum,

Bochum, Germany3 Department of Gastroenterology and Hepatology, Kliniken Bergmannsheil, University of

Bochum, Germany

Improved methods for the early diagnosis of colorectal cancer by way of sensitive and specifictumour markers are highly desirable. Therefore, efficient strategies for biomarker discovery areurgently needed. Here we present an approach that is based on the direct experimental access toproteins released by SW620 human colorectal cancer cells in vitro. A 2-D map and a catalogue ofthis subproteome – here termed the secretome – were established comprising more than 320identified proteins which translate into approximately 220 distinct genes. As the majority of thesecretome constituents were nominally cellular proteins, we directly compared the secretomeand the total proteome by 2-D-DIGE analysis. We provide evidence that unspecific releasethrough cell death, classical secretion, ectodomain shedding, and exosomal release contribute tothe secretome in vitro, presumably reflecting the mechanisms in vivo which lead to the occur-rence of tumour-specific proteins in the circulation. These data together with the fact that theSW620 secretome catalogue, as presented here, does comprise a large number of known andnovel biomarker candidates, validates our approach to isolate and characterize the tumour cellsecretome in vitro as a rich source for tumour biomarkers.

Received: July 7, 2006Accepted: August 25, 2006

Keywords:

Colon cancer / Exosome / Kallikrein-6 / Secretome / Soluble E-cadherin

Proteomics Clin. Appl. 2007, 1, 47–61 47

1 Introduction

Colorectal cancer is the second leading cause of cancer deathin the Western world and 200 000 people annually die from itin Europe [1]. Colorectal cancers can be cured when detected

early. However, 60% of patients present with locally advanceddisease or distant metastases, rendering treatment optionsless promising. The 5-year survival rate for patients withmetastatic disease is approximately 7% [2]. Screening mod-alities currently in clinical practice such as the faecal occultblood test and colonoscopy have already contributed to asubtle decline in mortality from colorectal cancer over thepast decade. However, they have severe limitations thatrestrict their applicability. The faecal occult blood test, theonly noninvasive screening test, has a low sensitivity, needsstrict dietary restrictions to reduce false positive results anddetects established cancer rather than adenomas.

Correspondence: Dr. Irmgard Schwarte-Waldhoff, Department ofInternal Medicine, IMBL, Knappschaftskrankenhaus, In derSchornau 23-25, D 44892 Bochum, GermanyE-mail: [email protected]: 149-234-299-4380

Abbreviations: APP, amyloid precursor protein; CEA, carcinoem-bryonic antigen; HCT, high capacity IT; PDI, protein disulphideisomerase; PGK, phosphoglycerate kinase * Both the authors have contributed equally to this work.

DOI 10.1002/prca.200600491

© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.clinical.proteomics-journal.com

48 H. C. Diehl et al. Proteomics Clin. Appl. 2007, 1, 47–61

Colonoscopy is regarded as the gold standard for detect-ing adenomas and early carcinomas. Major drawbacks, how-ever, are high costs and poor patient compliance. Thus, thequest is for blood-based biomarkers such as carcinoem-bryonic antigen (CEA). While very useful for monitoringmetastatic disease, CEA as a single marker does not reachhigh enough sensitivity for screening purposes. The ratio-nale for discovering additional novel diagnostic biomarkersresides in the goal to increase sensitivity and specificitythrough marker multiplexing.

The diagnostic potential generally ascribed to serum andplasma is reflected in the huge efforts put into their charac-terization world wide, such as the Plasma Proteome Projectorganized by the HUPO [3]. In a multilaboratory effort, anannotated database of normal human plasma proteins hasrecently been established comprising approximately 7500distinct proteins and isoforms which correspond to 3800unique genes. Multiple approaches to directly discover puta-tive tumour biomarkers within the plasma proteome of can-cer patients have been reported during the last few years.Due to the complexity and the enormous dynamic range ofthe plasma proteome, however, this is an extraordinary chal-lenge. It has been estimated that the range of protein con-centrations spans ten orders of magnitude between the mostand the least abundant proteins [4]. Proteins derived fromtumours and present in the circulation are expected withinthe latter class of the less abundant proteins. In line withthis, several attempts aimed at the identification of tumourbiomarkers in blood have apparently detected effects oftumours on abundant serum proteins rather than tumour-derived proteins per se [3, 5–8]. Breakdown products of serumproteins in cancer patient serum samples can be due to theactivity of tumour-expressed proteases. So, such indirectalterations in the serum-proteome can be tumour cellinduced, however, it appears less probable that they mayprove specific for distinct tumour entities and stages. Still,tumour-derived proteins are regarded as the most promisingbiomarkers.

Proteins released by human tumour cells in vivo andreaching the circulation are strongly outweighed by all thenormal blood constituents. Thus, seeking for an alternativesource for biomarker discovery, we have developed a protocolthat provides direct experimental access to this promisingsubproteome of proteins released by human colon cancercells in vitro. Release of proteins from tumour cells in vivoand in vitro is due to diverse mechanisms and is not confinedto classical secretion, although similar subproteomes forsake of simplicity have previously been and will in the fol-lowing be termed ‘secretome’. Classical secretion is the mostobvious mode of protein release and is expected to be rele-vant for proteins such as extracellular matrix molecules,proteases and protease inhibitors as well as growth factorsand cytokines. Also, the release of membrane proteinsthrough proteolytic shedding of ectodomains or throughcleavage of phospholipids membrane anchors as exemplifiedby the serum marker CEA [9–12], belongs to the well-known

mechanisms. Although hitherto less appreciated, the so-called exosomes represent another putative source of plasmabiomarkers [13]. Exosomes are membrane-coated vesiclesderived from multivesicular bodies in the late endosomalcompartment [14–16]. They have first been detected as pro-duced by immune cells and are regarded as important de-vices for intercellular communication involved in the reg-ulation of immune responses. Exosomes are now known toalso be produced by intestinal epithelial cells and by cancercells [13, 17]. Tumour-derived exosomes have previously beenfound in sera from cancer patients [18]. The composition ofexosomes has been investigated in some detail demonstrat-ing that they represent a distinct subproteome [16, 19–22].Lastly, biomarkers detectable in the patient plasma/serummay also simply be due to tumour tissue breakdown. Solidtumours often harbour large numbers of apoptotic cells.Also, aggressive and rapidly growing tumours, in particular,contain necrotic areas induced by hypoxia due to limitedblood supply. The fate of these apoptotic and necrotic tumourcells is incompletely understood. Engulfment by macro-phages or other scavenger cells may be involved in theirturnover. Tumour-derived undigested or partially digestedproteins may indirectly be released by macrophages in solu-ble form or in the form of exosomes.

The composition of the secretome in this wider senseappears to be rather complex. Thus, we have set up an em-pirical approach adequate to isolate, identify and characterizethe subset of proteins released by human carcinoma cells invitro regardless of the underlying mechanisms of release. Tothat aim, we have chosen the SW620 colon carcinoma cellline as a model. Proteins were harvested from conditionedmedia, concentrated and resolved using 2-DE and stainedwith an MS-compatible silver-staining procedure [23]. Twohundred and seventy-two out of 355 protein spots isolatedfrom a master gel could be identified through MALDI-TOF/TOF-MS translating into 178 distinct protein species.Another 40 proteins were identified after isolation fromother gels which could not be assigned to distinct spots in themaster gel. Together, these proteins represent one of themost comprehensive secretome catalogues published so far.

In addition, as the majority of the secretome constituentswere found to be nominally cellular proteins, we have di-rectly compared the cellular proteome and the secretomederived from the same cells through 2-D-DIGE analysis. Thisapproach illustrates that both compartments are distinct, al-though a certain degree of overlap is apparent. Examples forproteins present in both compartments and examples forsecretome-specific proteins were assayed and confirmed byWestern blot analysis. Similar results as obtained withSW620 cells were seen in a set of further human colorectalcancer cell lines validating our catalogue as an alternativesource for colorectal tumour biomarkers. Ultimately, we havealso addressed the mechanisms of protein release and pres-ent examples for classical secretion such as kallikrein-6, forectodomain shedding, i.e., soluble E-cadherin and for exoso-mal release such as syntenin.



2 Materials and methods

2.1 Cell culture, preparation of the secretome

for 2-D-gel electrophoresis and exosome

enrichment

The human colorectal carcinoma cell lines SW620 andSW480 were obtained from the American Type Culture Col-lection (Rockville, MD, USA), the cell lines HT29 and SW948were kindly provided by M. Strauss (Berlin). All cells weremaintained in DMEM supplemented with 10% FCS, 2 mMglutamine, 100 U/mL penicillin and 100 mg/mL streptomy-cin. For secretome preparation, cells were cultured in stand-ard medium until they reached a confluence of 60–70%.Cells were then washed three times with DMEM and incu-bated in serum-free medium with hydrocortisone at 1 ng/mL and ITS additives (Sigma, St. Louis, MO, USA) consist-ing of 5 mg/mL insulin, 5 mg/mL transferrin and 5 ng/mLsodium selenite for another 2 days. This protocol did notmeasurably influence the apoptosis rate compared to stand-ard culture conditions. The fraction of dead cells was deter-mined by trypane blue exclusion to be lower than 2%. Theconditioned medium was collected from the culture dishesand cooled down on ice. Floating cells and cellular debriswere removed by centrifugation (2006g, 10 min) followedby sterile filtration (pore size: 0.2 mm). Protease inhibitors(7 nm pepstatin, 85 mg/mL PMSF and Inhibitor CocktailComplete™ (Roche, Mannheim, Germany) were added andthe proteins were concentrated by ultrafiltration using Cen-triplus YM-3 centrifugal filter devices (Millipore, Bedford,MA, USA) according to the manufacturer’s instructions. Thetotal protein amount was determined using a standardBradford protein assay (BioRad, Hercules, CA, USA). For2-D-gel electrophoresis, the concentrated proteins weredesalted using Micro Bio-Spin® P-6 chromatography col-umns (BioRad), dried in a SpeedVac® and resuspended at afinal concentration of 10 mg/mL in IEF sample buffer (30 mMTris-HCl, 2 M thiourea, 7 M urea, 4% CHAPS; pH 8.5). DTT(1.08 g/mL, BioRad) and ampholine 2–4 (GE Healthcare,Freiburg, Germany) were added to the protein samples to afinal concentration of 75 mM and 2% v/v, respectively.

The enrichment of exosomes derived from the SW620secretome was performed as described by van Niel et al. [17]with modifications. In brief, secretome samples were subjectto ultracentrifugation at 100 0006g for 24 h at 47C using a T-865 Titanium rotor (SORVALL®, Langenselbold, Germany).The supernatant was recovered for separate analysis, thepellet was washed and resuspended in PBS.

2.2 2-DE and protein staining with silver or CyDyes

Carrier ampholyte (CA) based IEF was performed in a self-made IEF-chamber by running a voltage gradient for 21.3 has described [24]. The ejected tube gels (20 cm61.5 mm)were incubated in equilibration buffer (125 mM Tris, 40% w/v glycerol, 3% w/v SDS, 65 mM DTT; pH 6.8) for 10 min,

transferred to the second-dimension polyacrylamide gel(SDS-PAGE, 20 cm630 cm61.5 mm; 15.2% total acryl-amide, 1.3% bisacrylamide) and fixed with 1.0% w/v agarose(Merck, Darmstadt, Germany) containing 0.01% w/v Bro-mophenol blue dye (Riedel de Haën AG Seelze, Hannover,Germany). The gel run was performed in a Desaphor VA 300system [24]. For each preparative gel, 400 mg SW620 secre-tome protein was loaded. The gels were run in both dimen-sions followed by staining of the proteins with a MALDIcompatible silver-staining procedure [25]. The staining wasstopped when the less abundant spots appeared [23]. Gelswere scanned with a UMAX Mirage II scanner.

The secretome lysate comparison was performed usingthe 2-D-DIGE technology (GE Healthcare). CyDye stocks(1 nmol/mL) were diluted in anhydrous DMF p.a. (Sigma) toa final concentration of 400 pmol/mL and 400 pmol dye wasadded per 50 mg protein. The secretome was labelled withCy3 and the cellular proteins with Cy5. The samples werevortexed, centrifuged for 10 s, and incubated on ice for30 min in the dark. The labelling reaction was stopped by theaddition of 1 mL l-lysine stock solution (Sigma) (10 mM) per400 pmol dye. DIGE gels were scanned with a Typhoon 9400Variable Mode Imager (GE Healthcare). Excitation andemission wavelengths were chosen specifically for each ofthe dyes according to recommendations of the manufacturer.When proteins were isolated and identified from a DIGE gel,300 mg of unlabelled SW620 secretome protein was added tothe DIGE-stained samples.

For 2-D Western blotting the gel size was reduced (IEF,7 cm60.9 cm; SDS PAGE, 7 cm66 cm67 mm). Blottingwas performed as described below.

2.3 In-gel tryptic digestion and peptide analysis, by

MALDI-TOF/TOF-MS and nano-HPLC/ESI-MS/MS

for protein identification via bioinformatics

For protein identification, silver-stained and CyDye-labelledprotein spots were manually excised and in-gel digested withtrypsin (0.033 mg/mL, specific activity 500 U/mg, Promega,Mannheim, Germany) at 377C overnight. Peptides wereextracted from the gels as previously described for MALDI-TOF/TOF-MS analysis [26]. For nano-HPLC/ESI-MS/MSanalyses, peptides were extracted twice from gel pieces with50% ACN/2.5% formic acid (50:50).

Tryptic peptides were analysed with MALDI-TOF/TOF-MS using an Ultraflex II™ (Bruker Daltonics, Bremen, Ger-many) according to the manufacturer’s instructions. Pep-tides were spotted onto an MTP AnchorChip™ 800/384 TFtarget (Bruker Daltonics). The target positions were manu-ally coated with a saturated solution of CHCA matrix. Driedsamples were subsequently washed with 0.1% TFA toremove sodium and potassium adducts. The spectra wereacquired in the positive mode with a target voltage of 20 kVand a pulsed ion extraction of 17.25 kV. The reflector voltagewas set to 21 kV and detector voltage to 1.7 kV. Internal cali-bration of PMF spectra was performed using the autolysis



products of trypsin observed at m/z-values of 842.51, 1045.56and 2211.11 (monoisotopic masses are given). PMF spectrawere processed using the FlexAnalysis™ software (BrukerDaltonics). For subsequent protein identification the datawere sent to the ProteinScape™ database (Bruker Daltonics).Searches were started from ProteinScape database, using theProFound [27] or MASCOT [28] search algorithms. A Pro-Found score of .1.65 was set as threshold for protein iden-tification. The following search parameters were selected:fixed cysteine modification with propionamide, methionineoxidation as variable modification, one and two maximalmissed cleavage sites in case of incomplete trypsin hydro-lysis, mass tolerance of 50 and 100 ppm, respectively, an MWmass range from 5.0 to 250.0 kDa and pI range of 2.0 to 12.0.All searches were run in the human protein subdatabase ofthe NCBI (http://www.ncbi.nlm.nih.gov, June 2004–Novem-ber 2005). Peptides from protein spots with a low or not sig-nificant ProFound score were automatically selected for MS/MS using the MALDI-TOF/TOF instrument. The obtaineddata were assigned with the SEQUEST™ algorithm [29]. Thesame search parameters, as described above, were used withthe following exception: the peptide mass tolerance was setat 0.5 Da for monoisotopic masses and at 0.3 Da for frag-ment masses. A SEQUEST score of .1.5 for a single peptidewas set as threshold for protein identification. All searcheswere repeated using the same parameters except that thetaxonomy was extended to Mammalia, in order to identifyputative contaminating bovine proteins originating from theculture media.

MS/MS analyses of proteins isolated from the DIGE gelwere performed on a high capacity IT system (HCT plus,Bruker Daltonics) in conjunction with online capillaryHPLC. Online RP capillary HPLC separations using theDionex LC Packings HPLC systems (Dionex LC Packings,Idstein, Germany) was performed as previously described bySchaefer et al. [30]. The mass spectrometer was operated inthe sensitive mode with the following parameters: capillaryvoltage, 1400 V; end plate offset, 500 V; dry gas, 10.0 L/min;dry temperature, 1607C; aimed ion charge control, 150 000;maximal fill-time, 500 ms. The nano-ESI source (BrukerDaltonics) was equipped with distal coated SilicaTips (FS360-20-10-D; New Objective). MS spectra were the sum of sevenindividual scans ranging from m/z 300 to 1400 with a scan-ning speed of 8100 m/z?s–1. Data-dependent software (HCTplus, Esquire Controle, Bruker Daltonics) was employed toselect the two most intense, multiple-charged peptide ionsdetected within the MS spectra to subsequently perform MS/MS. To generate fragment ions, low energy CID was done onpreviously isolated peptide ions by applying a fragmentationamplitude of 0.6 V. Generally, MS/MS spectra were the sumof four scans ranging from m/z 100 to 2200 at a scan rate of26000 m/z?s–1. Exclusion limits were automatically placedon previously selected m/z for 1.2 min. The IT instrumentwas externally calibrated with commercially available stand-ard compounds. For protein identification, raw MS/MS datawere searched using the SEQUEST algorithm against the

NCBI protein sequence database restricted to Homo sapiens.The reliability of protein identification was checked manu-ally; two or more peptides with SEQUEST scores .3.0 had tobe assigned in order to consider protein identification assignificant.

2.4 Isolation of protein lysates and

Western blotting

Cells were lysed in NP-40 lysis buffer (25 mM Tris-HCl,pH 7.4, 0.5% NP-40, 100 mM NaCl, 1 mM EDTA) contain-ing a protease inhibitor cocktail (Roche) and 1 mM PMSF.Total protein lysates and secretomes of various colon carci-noma cell lines were subjected to standard 1-D-SDS-PAGE(12%) and transferred to a PVDF membrane (Immobilon™-FL, Millipore) by semidry blotting. Immunoblot blockingwas carried out in Odyssey blocking buffer (LI-COR Bio-sciences, Bad Homburg, Germany) and PBS (1:1) overnightat 4–107C. The membranes were incubated for 2 h with therespective antibodies: phosphoglycerate kinase (PGK; sheepantihuman PGK antiserum; 1:2000, kindly provided by P.Hogg, Australia), protein disulphide isomerase (PDI; mono-clonal mouse antihuman PDI; 1:1000, Affinity BioReagents,MA3-019), cystatin C (rabbit antihuman cystatin C anti-serum; 1:500 upstate, 06-458), E-cadherin (monoclonalmouse antihuman; 1:1000, Zymed, 13-1700), syntenin (rab-bit antihuman syntenin antiserum; 1:1000, Synaptic Sys-tems, 133002) and kallikrein-6 (rabbit antihuman kallikrein-6 antiserum; 1:500, Santa Cruz, SC-20624). After primaryantibody incubation blots were washed three times for10 min with PBS containing 0.05% Tween-20 (PBST). Fordetection, the immunoblots were incubated with the respec-tive secondary antibodies: donkey antisheep, goat antirabbitand goat antimouse, diluted 1:5000 in Odyssey buffer/PBS(1:1) solution. All secondary antibodies were coupled withthe fluorescent dye Alexa Flour 680. Signals were detectedusing the Odyssey Infrared Imaging System (LI-COR Bio-sciences).

3 Results and discussion

3.1 Preparative 2-D-gel electrophoresis of

the SW620 ‘secretome’

To establish a secretome catalogue of proteins releasedfrom cells in vitro, we have chosen the well-characterizedhuman SW620 cell line derived from a colorectal cancermetastasis. SW620 cells present an accumulation ofgenetic alterations characteristic of colorectal carcinomas:inactivated APC [31], p53 [32] and Smad4 [33] tumoursuppressor genes and an activated Ki-ras oncogene [34].SW620 cells were grown in standard medium with 10%fetal calf serum until reaching a confluence of approxi-mately 60%. Cells were then washed extensively to removeany residual serum-derived proteins, and were then cul-



tured for another 2 days in serum-free medium plus addi-tives as described [35]. Cell cultures under these condi-tions, like cultures in standard medium, routinely con-tained less than 2% dead cells as measured by trypane blueexclusion. Significant increases in the fraction of dead cellswere observed only after four and more days of culture inserum-free medium. Medium conditioned for 2 days washarvested and cleared from dead cells and from cellulardebris as described previously, and secretome proteinswere concentrated by ultrafiltration [35, 36]. We have cho-sen a gel-based approach in order to establish a reference2-D map of the colon cancer secretome. Proteins were runon high-resolution preparative 2-D gels and stained withsilver [23]. First, 355 protein spots were cut from a gelwhich served as a master (Fig. 1) and were subject to in-geltryptic digestion and MALDI-TOF/TOF-MS analysis. Totest for reproducibility of the protein pattern and toincrease the fraction of proteins identified, another threegels were run with independent secretome preparationsand protein patterns were matched with the master gel. Nomajor differences were detected between these gels, al-though, depending on slightly varying staining intensities,a number of additional spots were found. These additionalspots as well as a number of spots corresponding to themaster map were excised and subject to MALDI-TOF/TOF-MS analysis. In total, 272 of the 355 spots derived from themaster gel were identified (many of them from two ormore gels) (Supplementary Table 1), which translated into178 distinct proteins. Another 40 proteins were identifiedfrom the additional gels which could not be assigned todistinct protein spots in the master gel (SupplementaryTable 2). The additional search in Mammalia-identifiedbovine albumin in spots 268–271, only (data not shown).The scarcity of bovine proteins in the secretome underlinesthe effectiveness of our protocol to minimize mediumcontaminants. Of note, this secretome catalogue is pre-dominantly comprised of nominally cellular proteins,derived from cytoplasm and nuclear localizations and alsofrom the cytoskeleton, ER and mitochondrial matrix. Manynominally cellular proteins have previously been shown toserve as serum biomarkers in various diseases, in particu-lar, when increased tissue breakdown occurs. Basically, notonly tumour-specific proteins (i.e. oncofetal antigens) butalso aberrantly increased amounts of ‘normal’ proteins inplasma may indicate disease. Thus, we interrogated theHuman Plasma Proteome Database as a reference for pro-teins present in plasma of healthy humans (www.plasma-proteomedatabase.org) for all proteins identified in theSW620 secretome. In fact, we found either the same orclosely related proteins in .70% of the cases (compareSupplementary Table 1).

Protein release does not appear to be due to stressfulculture conditions. The secretome pattern derived from cul-tures showing increased apoptosis after 4 days of serumdepletion largely differed from the master map as describedhere (data not shown).

3.2 2-D DIGE analysis of the SW620 secretome versus

cell lysate

To get more insight into the nature of the secretome fraction,we wished to compare it to the SW620 cellular proteome andfor that aim, we made use of the 2-D-DIGE technology(Fig. 2a). The result shows some overlap of both proteomesas well as basic differences and illustrates a number of inter-esting findings: First, a larger number of protein spots aredetected in the lysate than in the secretome fraction, indi-cating that the cellular proteome is more complex than thesubset of released proteins. There are many cellular proteins,among them abundant proteins, which cannot be detected inthe secretome fraction; the secretome, thus, is not just areflection of the cellular proteome. For particular proteins,on the other hand, there is a significant overlap of proteinsfound in both fractions at similar relative concentrations.Their presence in the secretome may be due to unspecificrelease through cell death. Examples for this group of pro-teins were validated through Western blotting as described inthe next paragraph. Surprisingly, several nominally intracel-lular proteins like syntenin (see below) were even indicatedas highly enriched in the secretome fraction in this empiricapproach. The 2-D-DIGE analysis also revealed limitations ofour secretome catalogue: A substantial fraction of the proteinspots categorized as secretome-specific through the directsecretome versus lysate comparison had not been identified,some of them despite multiple attempts of MALDI-TOF/TOF-MS. In addition, several proteins are stained with thefluorescent dyes which were not visible on silver-stained gels.For example, chains of proteins not running as spots butstretching over a range of several kDa in the second dimen-sion – a pattern which is suggestive of glycoprotein’s – areexclusively visible on the DIGE gel. These appear to berestricted to the secretome fraction consistent with the factthat many secreted proteins are glycosylated. This category ofproteins will also be addressed below.

3.3 Validation of 2-D-DIGE results by Western

blotting

We have chosen some examples to validate the presence andthe relative concentrations of the identified proteins in theSW620 cell lysate and secretome, respectively. The sameamount of cell lysate and secretome protein was blotted forvalidation of proteins like PGK and PDI, which appear to bepresent at similar concentrations in both compartmentsaccording to the DIGE gel (Fig. 2b). The Western blots con-firmed that similar amounts of proteins, both PGK and PDI,are found in the SW620 cell lysate and the secretome(Fig. 2b). One of the most abundant secretome-specific pro-teins was cystatin C (CST3), and the 3-D view suggested itsexclusive presence in the secretome fraction. Therefore, ten-fold more lysate protein than secretome protein was blotted.Still, a much stronger cystatin C-specific signal was detectedin the secretome lane (Fig. 2b). We conclude that the



Figure 1. A master gel of the SW620 secretome to establish a 2-D master map. A secretome preparation derived from SW620 human col-orectal carcinoma cells was separated on a preparative 2-D gel. Proteins were stained with an MS compatible silver-staining procedure.Three hundred and fifty-five protein spots were excised as indicated. Identified proteins are listed in Supplementary Table 1. The stronglystained area in the upper middle of the gel, depicted by spots 263 and 264, is due to the medium supplement transferrin.



Figure 2. A direct comparison ofcell lysate vs. secretomethrough 2-D-DIGE and exemp-lary verification of resultsthrough Western blotting. (a) Acell lysate preparation derivedfrom SW620 cells and secre-tome preparations from thesame cells were compared on asingle DIGE gel. Lysate proteinsare shown in red, secretomeproteins are depicted in green.Three groups of proteins can bedetected: those found in bothcompartments which come upin (different tastes of) yellow, anumber of red spots derivedfrom the cell lysate and somesecretome-specific spots ingreen. Two examples of the firstgroup (PGK and PDI) and oneexample of the latter group(cystatin C, CST3) were furtheranalysed. (b) PGK, PDI andcystatin C are shown in coloursections as indicated with whiteboxes in (a). The 3-D views ofthe respective gel regions allowfor a more quantitative estimateof the respective protein con-centrations. The relative con-centration of the three proteinsin SW620 secretome and lysatewas validated through Westernblot analysis using specific anti-bodies as shown in the panelson the right. Same amounts ofsecretome and lysate proteinwere loaded for the PGK and PDIWestern blots (5 mg), whereas inthe case of cystatin C, ten-foldthe amount of lysate proteinwas loaded because no signalwas detected in the lysate on theDIGE gel. Still, the Western blotgives a much stronger signal inthe secretome lane, illustratingthe vast enrichment of cystatin Cin this compartment. Also indi-cated in (a) are some secretome-specific proteins which wereidentified in a targeted approach(compare Table 1). One of themis kallikrein-6, which runs as achain of signals extended in thesecond-dimension suggestiveof glycoproteins, as discussed inthe body of the text (compareFig. 4).



2-D-DIGE results provide qualitatively and quantitativelyvaluable data as they were reliably confirmed through West-ern blotting.

Next, we wished to analyse, if the expression and relativeconcentrations of these proteins in either the proteome orthe secretome fraction is a specific feature of SW620 cells orcan be extended to further colorectal carcinoma cell lines.Cell lysate and secretome protein fractions were preparedfrom the three additional human colorectal carcinoma celllines SW480, HT29 and SW948. SW480 is a cell line thatwas established from the primary tumour of the samepatient whose metastasis gave rise to the SW620 cell line,and like SW620 carries mutations in APC, Ki-ras, p53 andSmad4. HT29 and SW948 cells also have accumulated asimilar set of mutations except that HT29 harbours mutatedB-raf rather than Ki-ras and SW948 does express Smad4protein although in a presumably functionally inactive formthrough a missense mutation in exon 11 [33, 37–39]. Figure3 shows that all four human carcinoma cell lines gavesimilar results in Western blots, namely, similar proteinlevels of PGK and PDI in lysates and secretomes, respec-tively, and a huge enrichment of cystatin C in the secretomefraction (approximately 50-fold). The cystatin C proteins inthe cell lysates as detected on the Western blot have aslightly larger apparent MW which is presumably due to thepresence of the signal peptide, consistent with the fact thatcystatin C is secreted through the classical leader sequencedependent pathway.

In conclusion, this approach is adequate to decipherproteins secreted through the classical pathway. In addi-tion, the appearance of abundant cellular proteins in thefraction of released proteins also appears to be a commonfinding and may reflect unspecific release through cellularbreakdown. Putative biomarkers may also be recruitedfrom this category of proteins. Of note, PGK has previouslybeen isolated from conditioned media of tumour cells andhas been assigned an alternative novel function in blood asa plasmin reductase, which initiates the proteolytic releaseof the endogenous angiogenesis inhibitor angiostatin fromplasmin [40]. Importantly, increased plasma levels of PGKhave previously been reported in lung cancer patients andwere associated with poor prognosis [41]. PDI is an ubi-quitous, abundant protein that is located primarily in theER, where it functions as a molecular chaperone and cata-lyses the formation of disulphide bridges between specificcysteines upon maturation of extracellular proteins. PDIhas also been assigned an extracellular function: PDI issynonymous with prolyl 4-hydroxylase, which in the tetra-meric form catalyzes the formation of 4-hydroxyproline incollagen. Of note, a profiling approach of the cell surfaceproteome displayed an abundance of PDI along with anumber of other chaperones on the surface of tumour cells[42]. These findings illustrate that proteins often can befound at previously unsuspected localizations to performunsuspected functions. Cystatin C is among the mostabundant proteins in the secretome. Cystatins constitute a

Figure 3. Relative concentrations of PGK, PDI and cystatin C inthe lysates and secretomes of a set of human colorectal carci-noma cell lines. Validation through Western blot analysis ofproteins PGK, PDI and cystatin C was extended to the humancolorectal cancer cell lines SW480, HT29 and SW948, in com-parison to SW620. As in Fig. 2, same amounts of secretome (S)and lysate (L) protein were loaded for the PGK and PDI Westernblots (5 mg), whereas in the case of cystatin C (CST3), ten-foldthe amount of lysate protein was loaded. All cell lines showedsimilar results: relative concentrations of PGK and PDI weresimilar in the secretome and lysate fractions, whereas cystatinC was enriched in the secretome by a factor of approximately50-fold.

superfamily of cysteine protease inhibitors that collectivelyregulate a variety of biological processes including resis-tance to bacterial and viral infections, modulation ofimmune and inflammatory responses as well as tumourinvasion and metastasis. Cystatin C is the sole cysteine pro-tease inhibitor detected in this study. Cystatin C belongs tothe type 2 secretory cystatins, it is ubiquitously expressedand inhibits cysteine proteases of the cathepsin family ofwhich cathepsin B is the most prevalent member. Upontumour progression, cathepsins stimulate cell migration,invasion and metastasis through their role in extracellularmatrix remodelling and their function in proteolytic cas-cades [43, 44]. Altered serum concentrations of cystatin Cand of the cathepsin B/cystatin C complex were suggestedas diagnostic and prognostic indicators for diverse cancersincluding colorectal carcinomas [45].



3.4 A targeted approach to identify additional

secretome-specific proteins reveals

novel components released through classical

secretion and through ectodomain shedding

As discussed above, the 2-D-DIGE analysis illustrated sometechnical limitations of our approach: it revealed that thefraction of secretome-specific proteins successfully identifiedin our previous analysis appeared to be underrepresented.We wished to selectively increase this fraction through a tar-geted approach using the DIGE-derived assignment ofsecretome-specific spots. To that aim, a preparative DIGE-gelwas run with 50 mg, each, of the proteome and the secretomeproteins stained with the respective CyDyes plus 300 mg ofunstained secretome proteins. The fluorescence image wasprinted and the most obvious secretome-specific proteinswere determined in the gel (compare Fig. 2a), leading to theisolation of 12 protein spots from which nine could be iden-tified through nano-HPLC/ESI-MS/MS (Table 1). Three ofthese nine proteins were identified as transferrin and pre-sumably are breakdown products of the recombinant trans-ferrin used as medium additive in serum-free cultures. Theother six proteins represent an interesting extension of ourcatalogue.

Two very prominent protein spots were identified as beta-2-microglobulin. Beta-2-microglobulin can be detected onthe cell membrane of most cell types and is associated withhistocompatibility antigens in a noncovalent binding. Thefull-length beta-2-microglobulin sequence carries a signalsequence and a transmembrane region, but alternative spliceproducts devoid of the transmembrane region have also beenreported. It has been known for decades that beta-2-micro-globulin can be found in urine and in other body fluids.Recently, beta-2-microglobulin has also been detected in asecretome derived from pancreatic cancer cells where 46proteins were identified through multidimensional proteinidentification technology (MudPIT) [46].

A protein variably designated as Mac2-binding protein,galectin-3-binding protein or tumour associated antigen 90 K(TAA90K) was found as a spot within a predominant spotchain in the high molecular weight range (see Fig. 2a).TAA90K is a ubiquitously expressed secreted glycoprotein.Elevated concentrations were detected in the serum ofpatients with various types of breast, lung, endometrial,ovarian, and colorectal cancer [47, 48] and have been asso-ciated with poor prognosis.

The protein identified as kallikrein-6 was cut from aregion devoid of distinct protein spots but showing the abovementioned protein tracks of presumptive glycoproteins. Kal-likrein-6 is known to be a highly glycosylated secreted pro-tein. The assignment of kallikrein-6 to the 2-D protein pat-tern was addressed through Western blotting as shown in thenext paragraph.

Lastly, E-cadherin and amyloid-beta were detected in thisextension of secretome-specific proteins. Both are trans-membrane proteins, so their presence in the secretome frac-

tion at high abundance was surprising at a first glance. Theexperimental size of E-cadherin argued against the full-sizeprotein but suggested the presence of an E-cadherin frag-ment. This issue was also addressed through Western blot-ting of secretomes and lysates (see below).

3.5 The secreted serine protease kallikrein-6 is

abundantly expressed in colorectal cancer

secretomes

The protein identified as kallikrein-6 was selected as anexample for the above-mentioned presumptive glycoproteinswhich were visible on DIGE-gels but not on silver-stainedgels. Our hypothesis that this entire group of protein smearsextended in the second-dimension corresponded to kallik-rein-6 was addressed through 2-D-Western blotting using asmaller-sized gel and a kallikrein-6-specific antibody. In fact,a corresponding pattern was obtained (Figs. 4a and b). So,apparently, kallikrein-6 is secreted from SW620 cells in sig-nificant amounts. Kallikrein-6 expression was then assayedthrough 1-D Western blotting in our set of colorectal cancercells. Whereas kallikrein-6 expression is not detected inSW948 cells, SW480 and HT29 cells also express high levelsof kallikrein-6 (Fig. 4c).

Figure 4. 2-D pattern of kallikrein-6 as shown through 2-D West-ern blotting and kallikrein-6 expression in cell lines. The proteinidentified as kallikrein-6 was cut from one of the regions, whichon the DIGE-stained secretome show a chain of proteins exten-ded in the second dimension (a). 2-D Western blotting with a kal-likrein-6-specific antibody displayed a similar staining pattern (b)and thus confirmed that all of this protein chain is composed ofkallikrein-6. Expression in additional cell lines was assessed by1-D Western blotting (c). As for cystatin C 10-fold, the amount oflysate protein (5 mg secretome and 50 mg of lysate protein,respectively) was loaded because no lysate signal was detectedon the DIGE gel. Still, virtually no kallikrein-6-specific signalswere detected in the lysate samples.



Table 1. Proteins identified from the 2-D-DIGE gel by nano-HPLC/ESI-MS/MS

Protein NCBIaccession

Experimental Theoretical Sequencecoverage(%)

Peptides Chargestate

Sequestscore

pI MW(kDa)

pI MW(kDa)

Beta-2 microglobulin gi34616 6.7 15 5.5 12.8 28.2 IQVYSR 2 3.5VEHSDLSFSK 2 5VNHVTLSQPK 2 4.2WDRDM 2 1

Beta-2 microglobulin gi34617 6.7 15 5.5 12.8 23.6 IQVYSR 2 3.4VEHSDLSFSK 2 4.4VEHSDLSFSK 3 4.4VNHVTLSQPK 2 4

Kallikrein 6 gi30582543 6.9 32 7.5 26.9 4.5 LVHGGPCPAmDKa) 2 1.8ESSQEQSSVVR 2 5.1YTNWIQK 2 3.3

Transferrin gi31415705 5.8 52 6.7 77 7 SVIPSDGPSVACPAmVKa) 2 3.2DSGFQMNQLR 2 8.5DSGFQMOxNQLRb) 2 5.2DGAGDVAFVK 2 9.5IECPAmVSAETTEDCPAmIAKa) 2 4.4

Transferrin gi4389243 5.8 53 6.3 37.2 3 DSGFQMNQLR 2 5.3DSGFQMOxNQLRb) 2 5DGAGDVAFVK 2 9.2DSAHGFLK 2 3.8VPPR 1 1

Amyloid beta A4 protein,precursor, isoform b

gi41350939 4.5 85 4.7 84.8 16 LNMHMNVQNGK 2 6.9WDSDPSGTK 2 7.4FLHQER 2 2.8EVCPAmSEQAETGPCPAmRa) 2 6.2MSQVMR 2 5.3MSQVMOxRb) 2 1.8EWEEAER 2 2.4AVIQHFQEK 2 3.5VESLEQEAANER 2 8.7QQLVETHMAR 2 2VEAMLNDR 2 5.5VEAMOxLNDRb) 2 2.9HVFNMOxLKb) 2 3.3SQVMOxTHLRb) 2 2.8SQVMTHLR 2 4.2TEEISEVK 2 3MDAEFR 2 6.5

Transferrin gi62897069 4.2 85 6.8 77 9.6 SCPAmHTGLGRa) 2 1.3DGAGDVAFVK 2 9.1GDVAFVK 2 4.4IECPAmVSAETTEDCPAmIAKa) 2 3.8SCPAmHTAVGRa) 2 2.7HQTVPQNTGGK 2 4NPDPWAK 2 4.1DDTVCPAmLAKa) 2 2.9

E-cadherin gi6682963 4.2 . 85 4.6 99.7 1 DWVIPPISCPAmPENEKa) 2 1.6NLVQIK 1 3.4NLVQIK 2 2.5VTEPLDR 2 4.5VTEPLDRER 2 4.1NMOxFTINRb) 2 1.9



Table 1. Continued

Protein NCBIaccession

Experimental Theoretical Sequencecoverage(%)

Peptides Chargestate

Sequestscore

pI MW(kDa)

pI MW(kDa)

NMFTINR 2 3.4NTGVISVVTTGLDR 2 9.4TIFFCPAmERa) 2 1.5MALEVGDYK 2 6.5MOxALEVGDYKb) 2 6.6

90 K (Mac-2BP) gi483474 4.2 85 5.1 65.3 15.2 LADGGATNQGR 2 7.1VEIFYR 2 2.3STHTLDLSR 2 2.4YFYSR 2 1IDITLSSVK 2 6.1LASAYGAR 2 2.5SDLAVPSELALLK 2 1.9ASHEEVEGLVEK 2 6ASHEEVEGLVEK 3 1.9KSQLVYQSR 2 3.7SQLVYQSR 2 5.9TIAYENK 2 1.7

a) PAm, propionamide.b) Ox, oxidation.

The kallikreins constitute a subfamily of mammalianserine proteases which were first identified and most ofwhich are abundantly expressed in pancreatic fluids [49, 50].Prostate-specific antigen (PSA) is one of the human kallik-reins (hKH3), and is the most useful tumour marker forprostate cancer screening, diagnosis, prognosis and mon-itoring [51]. Kallikrein-6 (hKH6) has recently been suggestedas a new potential serum biomarker for a subset of ovariancancer patients carrying biologically aggressive tumours [52].An in silico expression analysis revealed up-regulation ofhKH6 not only in female genital tract cancers, but also ingastrointestinal cancers including those of the colon andpancreas [53]. Combined with our results of high abundanceof the kallikrein-6 protein in colorectal cancer cell secre-tomes, this may implicate hKH6 as a putative serum bio-marker in these cancers. Of note, kallikrein-6 in contrast tosome other kallikreins including PSA is not included in theHuman Plasma Proteome Database arguing for a cancer-specific origin.

The physiological role of kallikrein-6 is unknown. Intissue culture, the enzyme has been found to generateamyloidogenic fragments from the amyloid precursor pro-tein (APP), suggesting a potential for involvement in Alz-heimer’s disease [54]. Recombinant kallikrein-6 can alsocleave extracellular matrix proteins and a neutralizing anti-body reduced migration in a Boyden chamber assay [55, 56].Thus, kallikrein-6 may be causally involved in carcinogen-esis and may also constitute a target for therapeutic inter-vention.

3.6 Ectodomain shedding is the basis for the

occurrence of soluble E-cadherin and soluble

APP in colorectal cancer secretomes

E-cadherin is a 120 kDa transmembrane glycoprotein essen-tial for intercellular adhesion through the formation ofadherens junctions. Loss or reduced expression of E-cad-herin has been related to invasive behaviour in a wide rangeof carcinomas and E-cadherin is, thus, regarded as animportant invasion suppressor [57–60]. Here, we identifiedan abundant secretome-specific component of the SW620secretome as an 80 kDa form of the E-cadherin protein. Sol-uble forms of E-cadherin with an apparent molecular massof 80 kDa have previously been described, and represent thelarge N-terminal ectodomain fragment released throughproteolysis at transmembrane-proximal sites (sE-cadherin).All E-cadherin-specific peptides identified here throughnano-HPLC/ESI-MS/MS were derived from the N-terminalE-cadherin ectodomains. Thus, our finding is consistentwith E-cadherin ectodomain shedding in SW620 cells,resulting in rather high concentrations of the ectodomainfragment in the secretome.

The presence of full-length and of processed forms ofE-cadherin was assayed by Western blotting in the secretomesand lysates of colorectal cancer cell lines (Fig. 5). Of note,whereas rather high concentrations of soluble E-cadherinwere detected in the secretomes of SW620, HT29 and SW948cells, no full-length E-cadherin was present in the secre-tomes, but was readily detected in the corresponding cell



Figure 5. Detection of soluble E-cadherin in secretomes and offull-length E-cadherin in the corresponding lysates. E-cadherinexpression was assayed on Western blots loaded with 5 mg oflysate (L) and secretome (S) protein, respectively, of four differentcolon carcinoma cell lines. The anti-E-cadherin antibody readilydetects signals corresponding to the soluble form of E-cadherin(~80 kDa) in the secretomes, and to full-length E-cadherin(~120 kDa) in the lysates of SW620, HT29 and SW948 cells.SW480 express very low levels of E-cadherin, only.

lysates. SW480 cells are an exception in this context: they areknown to express very low levels of E-cadherin mRNA [61]; inline with this only trace amounts of sE-cadherin were detect-ed in the SW480 secretome.

Many proteins including cell-adhesion molecules,growth factor receptors and ligands are synthesized astransmembrane proteins and released from cells throughpost-translational proteolysis to yield bioactive molecules[62]. Soluble forms of E-cadherin have first been detected inconditioned media from MCF-7 human breast cancer cells[63]. Soluble E-cadherin is present in urine and serum fromhealthy individuals, where it may reflect shedding from theepithelium as part of the normal turnover of this molecule.Significantly increased levels of sE-cadherin have beenreported in urine and serum of cancer patients, presumablydue to tumour-associated proteolytic activities [64, 65].Diverse proteases including members of the MMP andADAM families, plasmin and presenilin-1/gamma-secretasehave been implicated in E-cadherin ectodomain shedding[66–69]. E-cadherin ectodomain shedding leads to dissocia-tion of E-cadherin, beta-catenin and alpha-catenin from thecytoskeleton. This process promotes disassembly of the E-cad-herin–catenin adhesion complex and induces cell migrationand cell scattering [67]. In addition, E-cadherin release canmodulate beta-catenin subcellular localization and down-stream signalling. In conclusion, production of (increasedlevels of) sE-cadherin by cancer cells can both represent a con-sequence of tumour-associated proteolytic activity, and alsopossess functional relevance with respect to cellular behaviourand expression profiles. To the best of our knowledge, it hasnot yet been addressed if sE-cadherin can serve as a serumbiomarker in colorectal cancer.

Another abundant secretome component identified inthe catalogue extension also appears to be released throughectodomain shedding, namely the APP. The APP geneencodes a widely expressed cell surface receptor and trans-membrane precursor protein that is cleaved by secretases torelease a soluble 100 000 kDa extracellular N-terminal frag-ment (sAPP), a small soluble peptide and a membrane-

bound C-terminal fragment. The large soluble N-terminalAPP fragment contains a Kunitz type proteinase inhibitordomain, it functions as a serine protease inhibitor and isanalogous to nexin II [70], a member of the family of sortingnexins, and to the platelet inhibitor of coagulation factor XI[71]. Interestingly, increased expression and processing ofAPP has recently been reported in pancreatic carcinoma andsAPP has been assigned a function to increase cellular pro-liferation in pancreatic cancer cells [72].

3.7 Syntenin exemplifies protein release through

exosome secretion

Following the direct comparison of lysate and secretomefractions through 2-D-DIGE, we also re-examined the pro-teins of our initial master map with respect to their relativedistribution in both compartments by comparative imageanalysis of the silver and the DIGE pattern. A survey of pro-teins highly enriched in the conditioned medium accordingto the DIGE analysis indicated syntenin, among others.Western blotting with a syntenin-specific antibody confirmedthis assignment, and indicated an approximately 20-foldhigher abundance of syntenin in the secretome fraction ascompared to cell lysates (Fig. 6a, note that different amountsof lysate and secretome protein were loaded). Syntenin-1 isprimarily regarded as an intracellular membrane-associatedmolecule. It is a PDZ (postsynaptic density protein-95, post-synaptic discs large and zona occludens-1) domain-contain-ing protein, which functions as an adaptor to bind cytoskel-etal proteins and signal transduction effectors.

Syntenin-1 also interacts with the cytoplasmic domain ofsyndecans, transmembrane glycoproteins carrying extra-cellular heparan sulphate moieties [73]. Syntenin-1 has alsobeen localized to the early secretory pathway (but is notsecreted itself), it is implicated in vesicular trafficking and itseems to be required for targeting TGF-alpha to the cell sur-face [74]. In a proteomic analysis of subcellular compart-ments prepared from polarized MDCK cells, syntenin-1 hasbeen assigned to apical endocytic vesicles [75]. All of thesedata do not explain the strong enrichment of syntenin in thesecretome fraction. To the best of our knowledge the singlepublished report about syntenin release from cells is theproteomic characterization of exosomes produced by den-dritic cells [16]. Exosomes are membrane coated vesicles firstdescribed in cells of hematopoietic origin and involved inantigen presentation. Exosomes can also be produced byintestinal epithelial cells and by cancer cells [13, 17]. Thus,we hypothesized that enrichment of syntenin in the secre-tome fraction might be due to exosomal release. To experi-mentally address this hypothesis we performed enrichmentof exosomes through ultracentrifugation [17]. Whereas noresidual syntenin signal could be detected in the supernatantafter ultracentrifugation, syntenin was significantly enrichedin the pellet as compared to the total secretome (Fig. 6b). Theenrichment of exosomes in the pellet was further confirmedby electron microscopy (data not shown). These results



Figure 6. Distribution of syntenin in proteome, secretome andexosome fractions. (a) The strong enrichment of syntenin in thesecretome vs. lysate fractions (approximately 20-fold) was con-firmed through Western blot analysis with a syntenin-specificantibody. Note that 8 mg of secretome (S) and 100 mg of lysate (L)protein were loaded, respectively. (b) After enrichment of exo-somes from the SW620 secretome through ultracentifugation,the syntenin-specific signal was strongly enhanced in the pelletfraction but virtually absent in the supernatant. Approximately4 mg protein was loaded in each lane. The same samples were runin parallel and were stained with silver to control for equal pro-tein loading (lower panel in b).

suggest that SW620 cells release exosomes into the super-natant and that these exosomes are the exclusive source ofsyntenin in SW620 conditioned medium. Ultimately, weexamined the 2-D-protein pattern from an exosome-en-riched fraction. Although the quality of the DIGE-picturewas unsatisfactory due to limiting amounts of protein (datanot shown), we could assign in this proteome a number ofadditional proteins previously identified in the secretome(Table 2) including cytoskeletal proteins such as actin, cofi-lin and profilin and major metabolic enzymes such as eno-lase, aldolase and PGK. Different from syntenin, however,exosomes were not the exclusive source of these proteins inthe SW620 secretome as observed by Western blot analysisof exosome enriched and exosome depleted secretome frac-tions with stratifin and PGK-specific antibodies (data notshown).

4 Concluding remarks

Here we have presented a catalogue of proteins released bycolorectal cancer cells in vitro as an alternative source forbiomarker discovery. This subproteome in a proteomic viewis tentatively called the secretome. Whereas a large fractionof the secretome is comprised of nominally cellular proteinsa direct comparison of the cellular proteome and the secre-tome demonstrated some overlap but very pronounced dif-ferences between the two compartments. We have exempla-rily addressed the mechanisms of protein release and pro-vided evidence that diverse mechanisms including classicalsecretion, ectodomain shedding and exosomal release con-tribute to the secretome. Validating our approach, this cata-logue comprises a number of proteins previously suspected

Table 2. Exosomal proteins as identified in the SW620 secretome

Protein NCBI accession Spot no. in master gel

Aldolase A gi229674 188/189ACTB protein gi15277503 161–166Heat shock 70 kDa protein 5 (glucose-regulated protein, 78 kDa) gi16507237 283–285Cofilin 1 (nonmuscle) gi15126676 41/42Cyclophilin A gi3659980 37–40Enolase 1 gi13325287 179Glyceraldehyde-3-phosphate dehydrogenase gi31645 129–131Heat shock 70 kDa protein 8 gi5729877 278–281Peroxiredoxin 1 gi32455266 44/45/47Peroxiredoxin 6 gi77744395 109PGK 1 gi4505763 190/191/193/195/334Platelet profilin gi5542166 28–3114-3-3 epsilon gi5803225 322/32514-3-3 gamma gi21464101 348/350/353/354Syntenin gi2795863 124Triosephosphate isomerase gi999892 104–106Tubulin beta gi897763 221



and/or verified as tumour biomarkers. Thus, we believe thatthe secretome represents a highly enriched source for puta-tive tumour biomarkers.

It will need further analyses to identify the mostpromising candidates therein. For example, the compar-ison of secretomes derived from different colorectal can-cer cell lines will indicate the most consistent secretomeconstituents establishing a ‘master secretome’ for thistumour entity. More importantly, differential analysis ofsecretomes in vitro will be an important extension of thiswork. It can be performed with tumour cells derived fromother tumour entities, with tumour cells of various dif-ferentiation grades and also for example with cell linesderived from colorectal adenomas. In this way, differentialsecretomics may provide evidence for putative stage- anddifferentiation-specific markers and help in the preselec-tion of serum biomarker candidates to be evaluated inpatient serum samples through more targeted approa-ches, such as immunoassays. Lastly, further enrichmentof specific subcompartments or subsets of the tumoursecretomes such as glycoproteins or exosomal proteinsand the use of gel-free methods for protein separationwill not only expand the secretome catalogue as a basisfor biomarker discovery, but will also improve our insightinto tumour cell biology.

This study was supported by grants from the DFG andDeutsche Krebshilfe – Dr. Mildred Scheel-Stiftung to I.S.-W. andW.S., by the BMBF and the NRW MWF to K.S. and H.E.M.and from the Ruhr-Universität Bochum to I.S.-W. and K.S. Theauthors gratefully acknowledge the contributions of B. Warscheidand S. Wiese for assistance with operating the HCT plus IT sys-tem and S. Hoppe, C. Bieling, S. Götze, E. Hawranke and K.Pfeiffer for excellent technical assistance. We also thank C. Ste-phan for his help with bioinformatics protein data analysis. PhilJ. Hogg (Sydney, Australia) is gratefully acknowledged for pro-viding PGK antiserum.

5 References

[1] Boyle, P., Ferlay, J., Ann. Oncol. 2005, 16, 481–488.

[2] Etzioni, R., Urban, N., Ramsey, S., McIntosh, M. et al., Nat.Rev. Cancer 2003, 3, 243–252.

[3] Omenn, G. S., States, D. J., Adamski, M., Blackwell, T. W. etal., Proteomics 2005, 5, 3226–3245.

[4] Anderson, N. L., Anderson, N. G., Mol. Cell. Proteomics 2002,1, 845–867.

[5] Conrads, T. P., Hood, B. L., Petricoin, E. F., III, Liotta, L. A. et al.,Expert Rev. Proteomics 2005, 2, 693–703.

[6] Jacobs, J. M., Adkins, J. N., Qian, W. J., Liu, T. et al., J. Pro-teome Res. 2005, 4, 1073–1085.

[7] Diamandis, E. P., J. Natl. Cancer Inst. 2004, 96, 353–356.

[8] Diamandis, E. P., Mol. Cell. Proteomics 2004, 3, 367–378.

[9] Hefta, S. A., Hefta, L. J., Lee, T. D., Paxton, R. J. et al., Proc.Natl. Acad. Sci. USA 1988, 85, 4648–4652.

[10] Sack, T. L., Gum, J. R., Low, M. G., Kim, Y. S., J. Clin. Invest.1988, 82, 586–593.

[11] Takami, N., Misumi, Y., Kuroki, M., Matsuoka, Y. et al., J. Biol.Chem. 1988, 263, 12716–12720.

[12] Jean, F., Malapert, P., Rougon, G., Barbet, J., Biochem. Bio-phys. Res. Commun. 1988, 155, 794–800.

[13] Wolfers, J., Lozier, A., Raposo, G., Regnault, A. et al., Nat.Med. 2001, 7, 297–303.

[14] Raposo, G., Nijman, H. W., Stoorvogel, W., Liejendekker, R.et al., J. Exp. Med. 1996, 183, 1161–1172.

[15] Escola, J. M., Kleijmeer, M. J., Stoorvogel, W., Griffith, J. M.et al., J. Biol. Chem. 1998, 273, 20121–20127.

[16] Thery, C., Boussac, M., Veron, P., Ricciardi-Castagnoli, P. etal., J. Immunol. 2001, 166, 7309–7318.

[17] van Niel, G., Raposo, G., Candalh, C., Boussac, M. et al.,Gastroenterology 2001, 121, 337–349.

[18] Caby, M. P., Lankar, D., Vincendeau-Scherrer, C., Raposo, G.et al., Int. Immunol. 2005, 17, 879–887.

[19] Hegmans, J. P., Bard, M. P., Hemmes, A., Luider, T. M. et al.,Am. J. Pathol. 2004, 164, 1807–1815.

[20] Mears, R., Craven, R. A., Hanrahan, S., Totty, N. et al., Prote-omics 2004, 4, 4019–4031.

[21] Pisitkun, T., Shen, R. F., Knepper, M. A., Proc. Natl. Acad. Sci.USA 2004, 101, 13368–13373.

[22] Thery, C., Zitvogel, L., Amigorena, S., Nat. Rev. Immunol.2002, 2, 569–579.

[23] Nesterenko, M. V., Tilley, M., Upton, S. J., J. Biochem. Bio-phys. Methods 1994, 28, 239–242.

[24] Klose, J., Kobalz, U., Electrophoresis 1995, 16, 1034–1059.

[25] Heukeshoven, J., Dernick, R., Electrophoresis 1988, 9, 28–32.

[26] Marcus, K., Immler, D., Sternberger, J., Meyer, H. E., Elec-trophoresis 2000, 21, 2622–2636.

[27] Zhang, W., Chait, B. T., Anal. Chem. 2000, 72, 2482–2489.

[28] Perkins, D. N., Pappin, D. J., Creasy, D. M., Cottrell, J. S.,Electrophoresis 1999, 20, 3551–3567.

[29] Yates, J. R., III, Eng, J. K., McCormack, A. L., Schieltz, D.,Anal. Chem. 1995, 67, 1426–1436.

[30] Schaefer, H., Chervet, J. P., Bunse, C., Joppich, C. et al., Pro-teomics 2004, 4, 2541–2544.

[31] Groden, J., Joslyn, G., Samowitz, W., Jones, D. et al., CancerRes. 1995, 55, 1531–1539.

[32] Rodrigues, N. R., Rowan, A., Smith, M. E., Kerr, I. B. et al.,Proc. Natl. Acad. Sci. USA 1990, 87, 7555–7559.

[33] Woodford-Richens, K. L., Rowan, A. J., Gorman, P., Halford,S. et al., Proc. Natl. Acad. Sci. USA 2001, 98, 9719–9723.

[34] Capon, D. J., Seeburg, P. H., McGrath, J. P., Hayflick, J. S. etal., Nature 1983, 304, 507–513.

[35] Volmer, M. W., Radacz, Y., Hahn, S. A., Klein-Scory, S. et al.,Proteomics 2004, 4, 1324–1334.

[36] Volmer, M. W., Stuhler, K., Zapatka, M., Schoneck, A. et al.,Proteomics 2005, 5, 2587–2601.

[37] Suter, C. M., Norrie, M., Ku, S. L., Cheong, K. F. et al., Br. J.Cancer 2003, 88, 413–419.

[38] Gayet, J., Zhou, X. P., Duval, A., Rolland, S. et al., Oncogene2001, 20, 5025–5032.



[39] Davies, H., Bignell, G. R., Cox, C., Stephens, P. et al., Nature2002, 417, 949–954.

[40] Lay, A. J., Jiang, X. M., Kisker, O., Flynn, E. et al., Nature2000, 408, 869–873.

[41] Chen, G., Gharib, T. G., Wang, H., Huang, C. C. et al., Proc.Natl. Acad. Sci. USA 2003, 100, 13537–13542.

[42] Shin, B. K., Wang, H., Yim, A. M., Le Naour, F. et al., J. Biol.Chem. 2003, 278, 7607–7616.

[43] Jedeszko, C., Sloane, B. F., Biol. Chem. 2004, 385, 1017–1027.

[44] Turk, V., Kos, J., Turk, B., Cancer Cell 2004, 5, 409–410.

[45] Zore, I., Krasovec, M., Cimerman, N., Kuhelj, R. et al., Biol.Chem. 2001, 382, 805–810.

[46] Mauri, P., Scarpa, A., Nascimbeni, A. C., Benazzi, L. et al.,FASEB J 2005, 19, 1125–1127.

[47] Grassadonia, A., Tinari, N., Iurisci, I., Piccolo, E. et al., Gly-coconj. J. 2004, 19, 551–556.

[48] Ulmer, T. A., Keeler, V., Loh, L., Chibbar, R. et al., J. Cell Bio-chem. 2006, 98, 1351–1366

[49] Borgono, C. A., Diamandis, E. P., Nat. Rev. Cancer 2004, 4,876–890.

[50] Yousef, G. M., Diamandis, E. P., Clin. Biochem. 2003, 36, 443–452.

[51] Obiezu, C. V., Diamandis, E. P., Cancer Lett. 2005, 224, 1–22.

[52] Anisowicz, A., Sotiropoulou, G., Stenman, G., Mok, S. C. etal., Mol. Med. 1996, 2, 624–636.

[53] Yousef, G. M., Borgono, C. A., White, N. M., Robb, J. D. et al.,Tumour Biol. 2004, 25, 282–289.

[54] Little, S. P., Dixon, E. P., Norris, F., Buckley, W. et al., J. Biol.Chem. 1997, 272, 25135–25142.

[55] Magklara, A., Mellati, A. A., Wasney, G. A., Little, S. P. et al.,Biochem. Biophys. Res. Commun. 2003, 307, 948–955.

[56] Ghosh, M. C., Grass, L., Soosaipillai, A., Sotiropoulou, G. etal., Tumour Biol. 2004, 25, 193–199.

[57] Behrens, J., Cancer Metastasis Rev. 1999, 18, 15–30.

[58] Christofori, G., Semb, H., Trends Biochem. Sci. 1999, 24, 73–76.

[59] Nollet, F., Berx, G., van Roy, F., Mol. Cell Biol. Res. Commun.1999, 2, 77–85.

[60] Van Aken, E., De Wever, O., Correia da Rocha, A. S., Mareel,M., Virchows Arch. 2001, 439, 725–751.

[61] Muller, N., Reinacher-Schick, A., Baldus, S., van Hengel, J. etal., Oncogene 2002, 21, 6049–6058.

[62] Hooper, N. M., Karran, E. H., Turner, A. J., Biochem. J. 1997,321, 265–279.

[63] Damsky, C. H., Richa, J., Solter, D., Knudsen, K. et al., Cell1983, 34, 455–466.

[64] Nawrocki-Raby, B., Gilles, C., Polette, M., Bruyneel, E. et al.,Int. J. Cancer 2003, 105, 790–795.

[65] Katayama, M., Hirai, S., Kamihagi, K., Nakagawa, K. et al., Br.J. Cancer 1994, 69, 580–585.

[66] Hayashido, Y., Hamana, T., Yoshioka, Y., Kitano, H. et al., Int.J. Oncol. 2005, 27, 693–698.

[67] Davies, G., Jiang, W. G., Mason, M. D., Clin. Cancer Res.2001, 7, 3289–3297.

[68] Marambaud, P., Shioi, J., Serban, G., Georgakopoulos, A. etal., EMBO J 2002, 21, 1948–1956.

[69] Maretzky, T., Reiss, K., Ludwig, A., Buchholz, J. et al., Proc.Natl. Acad. Sci. USA 2005, 102, 9182–9187.

[70] Van Nostrand, W. E., Wagner, S. L., Suzuki, M., Choi, B. H. etal., Nature 1989, 341, 546–549.

[71] Smith, R. P., Higuchi, D. A., Broze, G. J., Jr., Science 1990,248, 1126–1128.

[72] Hansel, D. E., Rahman, A., Wehner, S., Herzog, V. et al., Can-cer Res. 2003, 63, 7032–7037.

[73] Grootjans, J. J., Zimmermann, P., Reekmans, G., Smets, A.et al., Proc. Natl. Acad. Sci. USA 1997, 94, 13683–13688.

[74] Fernandez-Larrea, J., Merlos-Suarez, A., Urena, J. M., Base-lga, J. et al., Mol. Cell 1999, 3, 423–433.

[75] Fialka, I., Steinlein, P., Ahorn, H., Bock, G. et al., J. Biol.Chem. 1999, 274, 26233–26269.


A catalogue of proteins released by colorectal cancer cellsin vitro as an alternative source for biomarker discovery

Documents