J O U R N A L O F P R O T E O M I C S 9 1 ( 2 0 1 3 ) 8 4 – 9 6

Ava i l ab l e on l i ne a t www.sc i enced i r ec t . com

ScienceDirect

www.e l sev i e r . com/ loca te / j p ro t

The secretome of colon cancer stem cells containsdrug-metabolizing enzymes

Benjamin L Emminka, André Verheema, Winan J. Van Houdta, Ernst J.A. Stellera,Klaas M. Govaerta, Thang V. Phamb, Sander R. Piersmab, Inne H.M. Borel Rinkesa,Connie R. Jimenezb,⁎, Onno Kranenburga,⁎⁎aDepartment of Surgery, University Medical Center Utrecht, G04-228, PO Box 85500, 3508GA Utrecht, The NetherlandsbOncoproteomics Laboratory, Department of Medical Oncology, VU Medical Center, De Boelelaan 1117, 1081HV Amsterdam, The Netherlands

A R T I C L E I N F O

⁎ Correspondence to: C.R. Jimenez, OncoProteomVU University Medical Center, De Boelelaan 11⁎⁎ Correspondence to: O. Kranenburg, UniversiThe Netherlands. Tel.: +31 88 7558632; fax: +

E-mail addresses: c.jimenez@vumc.nl (C.R

1874-3919/$ – see front matter © 2013 Elseviehttp://dx.doi.org/10.1016/j.jprot.2013.06.027

A B S T R A C T

Article history:Received 20 April 2013Accepted 24 June 2013Available online 5 July 2013

Drug-resistant cancer stem cells (CSCs) have been implicated in tumor recurrence followingchemotherapy. However, the contribution of CSCs to drug-resistance in colorectal cancer isunclear and CSC-intrinsic drug-resistance mechanisms are ill-defined. Here, we addressthese issues by proteomic analysis of the secretomes of CSCs and isogenic differentiatedtumor cells (DTCs) isolated from three distinct metastasized colon tumors.Mass spectrometry-based proteomics identified 1254 unique proteins in the conditionedmedia of the paired CSC and DTC cultures. Ingenuity Pathway Analysis revealed thatproteins governing ‘Cell Death’ were most significantly enriched in the CSC secretome. Thevast majority of these (37/43) promote cell survival. The CSC secretome is also characterizedby a pro-survival Nrf2 antioxidant signature. Interestingly, proteome-maintenance net-works are highly enriched in the CSC secretome. CSCs also secrete high levels ofdrug-metabolizing enzymes, including aldehyde dehydrogenase 1 (ALDH1A1) and bleomycinhydrolase (BLMH). We show that these enzymes cause extracellular detoxification ofmaphosphamide and bleomycin respectively.We conclude that colorectal CSCs are characterized by extensive survival and anti-oxidantnetworks, which are likely to contribute to CSC-intrinsic drug-resistance. In addition, CSCsmaymodulate drug responses in nearby tumor cells by detoxifying chemotherapeutic drugsin the extracellular space.

Biological significanceCancer stem cells are thought to play an important role inmediating drug resistance and tumorrecurrence following chemotherapy. Therefore, it is important to identify the factors that aresecreted by them. Our results provide novel insights into the pathways that govern the intrinsicresistanceofCSCs to chemotherapyand, furthermore, demonstrate that they canalso inactivatechemotherapeutic drugs in the extracellular space. A better understanding of the pathways thatgovern drug resistance in CSCs may help in developing effective CSC-targeting drugs.

© 2013 Elsevier B.V. All rights reserved.

Keywords:Cancer stem cellsSecretomeDrug resistanceColorectal cancerMass-spectrometry

ics Laboratory, CCA 1-60, Department of Medical Oncology, VUmc-Cancer Center Amsterdam,17, 1081 HV Amsterdam, The Netherlands. Tel.: +31 20 4442340.ty Medical Center Utrecht, Department of Surgery, G04-228, PO Box 85500, 3508 GA, Utrecht,31 30 2541944.. Jimenez), o.kranenburg@umcutrecht.nl (O. Kranenburg).

r B.V. All rights reserved.

85J O U R N A L O F P R O T E O M I C S 9 1 ( 2 0 1 3 ) 8 4 – 9 6

1. Introduction

Colorectal cancer (CRC) is one of the leading causes of cancer-related deaths worldwide [1]. In recent years, the prognosis ofpatients with advanced disease has improved, mainly due to theaddition of new targeted drugs to standard treatment regimens[2–5]. However, CRC is characterized by highly divergent tumorresponses, and primary and acquired drug resistance is frequent-ly observed.

Colorectal tumors are hierarchically organized with a minorcancer stem cell (CSC) pool driving the formation of moredifferentiated less malignant cells making up the bulk of thetumor [6–9]. CSCs have also been implicated in drug resistanceand tumor recurrence although the nature of this relationship isonly beginning to be clarified. Colorectal CSCs display relativeresistance to oxaliplatin and 5-fluorouracil [10–12]. We recentlyidentified BIRC6, an Inhibitor of Apoptosis Protein (IAP) as aCSC-intrinsic factor that contributes to resistance against plati-numdrugs [12]. Likewise, CSCs from other tumor types appear tobe generally drug resistant [13]. Treatment of mice bearing

Sph= colonosphere culture enriched for CSCsDiff = isogenic differenated progeny of colonosphere culture

Sph Sph SphDiff Diff DiffL145 L146 L167A B

L1

DC

L167 Sph

L167 Diff

L146 Sph

L146 Diff

L145 Sph

L145 Diff

Fig. 1 – Secretome analysis of CSC-cultures and isogenic differentloaded with protein samples from CSC enriched cultures (Sph), acolorectal liver metastases: L145, L146 and L167. This gel was usproteins detected in CSC cultures and differentiated TCs. (C.) Heaproteins. (D.) Heatmap of supervised cluster analysis, only includSupplemental Table 1) and differentiated cultures.

colorectal tumor xenografts can cause a dramatic enrichment ofCSCs in post-therapy tumor tissue [10,14]. Likewise, post-treatment tumors in breast cancer patients are also characterizedby high CSC numbers, suggesting selective CSC-intrinsic drugresistance [15,16]. Nevertheless, the extent of chemotherapy-induced ‘selection’ for CSCs varies dramatically between studies,indicating that other mechanisms must play a role. Indeed,resistance of colorectal tumors to Irinotecan is mediated bydifferentiated cells expressing the drug efflux pump ABCB1 [17].This keeps intra-tumor drug concentrations low and allows CSCsand non-CSCs to survive treatment.

Embryonic SCs contain several genome protection mech-anisms, including high expression of DNA repair and anti-oxidant enzymes [18]. Breast CSCs also express high levels ofanti-oxidant enzymes that help prevent DNA damage andconsequent CSC ‘exhaustion’ [19]. Furthermore, multiple (C)SC populations, including colorectal CSCs, are characterizedby a low proliferation rate, which limits replication-associatedDNA damage [20]. Normal SCs of the intestine are anexception as they are actively cycling.

45

L167

L146

194 127 135

547

61 80

110

L167 Sph

L167 Diff

L146 Sph

L146 Diff

L145 Sph

L145 Diff

iated tumor cells. (A.) Coomassie-stained protein gradient gelnd differentiated tumor cells (Diff) from three distincted for mass-spectrometry analysis. (B.) Venn diagram of alltmap of unsupervised cluster analysis of all identifieding proteins enriched in CSC cultures (Table 1 and

Table 1 – Proteins enriched in the secretomes of cancer stem cell (CSC) cultures were compared to the secretomes of paired differentiated tumor cell cultures. Three tumorswere analyzed (L145, L146, L167). Proteins are categorized according to the average fold-enrichment in CSC secretomes with a cut-off at 2-fold. All proteins showed ≥1.5-foldenrichment in all 3 single comparisons. The p-values and fold change are based on the paired beta-binomial test. SecretomeP, NN-score above 0,5 indicates possiblenon-classical secretion, classical secreted proteins (having signaling peptide) are marked with *. Predicted exosome-derived proteins identified by the Exocarta database aremarked.

Protein Gene Accessionnumber

Spectral counts identified Foldchange

p-Value Location SecretomePNN-score

Human entriesExocarta 4.1

L145 L146 L167

Diff CSC Diff CSC Diff CSC

Chaperonin containing TCP1, subunit 7 (eta) CCT7 IPI00018465 0 6 0 5 0 4 ∞ 0.000 Cytoplasm 0.382 xFc fragment of IgG binding protein FCGBP IPI00242956 0 2 0 19 0 7 ∞ 0.000 Extracell space 0.706* xFibroblast growth factor 2 (basic) FGF2 IPI00154603 0 3 0 2 0 4 ∞ 0.002 Extracell space 0.239Fumarate hydratase FH IPI00296053 0 2 0 2 0 3 ∞ 0.005 Cytoplasm 0.493 xGlutathione synthetase GSS IPI00010706 0 5 0 2 0 4 ∞ 0.001 Cytoplasm 0.484 xN-sulfoglucosamine sulfohydrolase SGSH IPI00019988 0 4 0 2 0 7 ∞ 0.001 Cytoplasm 0.808*Proteasome (prosome, macropain) subunit,alpha type, 6

PSMA6 IPI00029623 0 3 0 3 0 3 ∞ 0.002 Cytoplasm 0.429 x

Ribosomal protein L23a RPL23A IPI00021266 0 3 0 4 0 3 ∞ 0.001 Cytoplasm 0.418T-complex 1 TCP1 IPI00290566 0 7 0 4 0 5 ∞ 0.000 Cytoplasm 0.518 xTransferrin TF IPI00022463 0 32 0 12 0 46 ∞ 0.000 Extracell space 0.478* xWD repeat domain 1 WDR1 IPI00746165 0 6 0 4 0 9 ∞ 0.000 Extracell space 0.518 xPhosphoglucomutase 1 PGM1 IPI00844159 3 10 0 10 0 28 15.3 0.000 Cytoplasm 0.43 xFerritin, light polypeptide FTL IPI00852596 0 11 0 9 2 4 14.7 0.001 Cytoplasm 0.383 xChaperonin containing TCP1, subunit 5 (epsilon) CCT5 IPI00010720 2 11 0 6 0 7 12.3 0.001 Cytoplasm 0.425 xAldehyde dehydrogenase 1 family, member A1 ALDH1A1 IPI00218914 0 10 2 14 2 22 11.7 0.000 Cytoplasm 0.501 xNAD synthetase 1 NADSYN1 IPI00306689 0 3 0 3 2 11 8.8 0.005 Cytoplasm 0.444Discoidin domain receptor tyrosine kinase 1 DDR1 IPI00001477 2 5 0 4 0 5 6.8 0.010 Plasma membr 0.094* xNeurolysin (metallopeptidase M3 family) NLN IPI00010346 3 6 0 2 0 12 6.4 0.008 Cytoplasm 0.450*Ribosomal protein S13 RPS13 IPI00221089 2 3 0 5 0 5 6.2 0.017 Unknown 0.816 xOlfactomedin 4 OLFM4 IPI00022255 0 5 12 36 0 31 5.6 0.000 Extracell space 0.649* xSerine/arginine-rich splicing factor 1 SRSF1 IPI00215884 0 6 0 4 3 5 5.6 0.016 Extracell space 0.919*Malic enzyme 1, NADP(+)-dependent, cytosolic ME1 IPI00008215 5 17 2 8 0 12 5.3 0.003 Cytoplasm 0.378 x

86JO

UR

NA

LO

FPR

OT

EO

MIC

S91

(2013)

84–96

EPH receptor B4 EPHB4 IPI00289342 0 2 0 3 3 9 5 0.018 Plasma membr 0.366* xSerpin peptidase inhibitor, clade B(ovalbumin), member 6

SERPINB6 IPI00413451 5 20 0 11 5 12 4.8 0.006 Cytoplasm 0.44 x

Leucyl/cystinyl aminopeptidase LNPEP IPI00221240 5 12 0 11 3 8 4.2 0.015 Cytoplasm 0.63CNDP dipeptidase 2 (metallopeptidaseM20 family)

CNDP2 IPI00177728 2 4 0 2 0 3 4.1 0.050 Cytoplasm 0.443 x

LSM4 homolog, U6 small nuclear RNAassociated (S, cerevisiae)

LSM4 IPI00294955 0 2 2 4 0 3 4 0.054 Nucleus 0.426

Ribosomal protein L23 RPL23 IPI00010153 2 3 0 2 0 3 3.6 0.078 Cytoplasm 0.582* xGlutathione peroxidase 2 (gastrointestinal) GPX2 IPI00298176 0 7 3 6 2 3 3.6 0.066 Nucleus 0.233Angiogenin, ribonuclease, RNase A family, 5 ANG IPI00008554 0 6 3 8 4 8 3.5 0.024 Extracell space 0.889* xN-acylaminoacyl-peptide hydrolase APEH IPI00337741 8 12 5 10 0 28 3.4 0.016 Nucleus 0.577 xGTP cyclohydrolase I feedback regulator GCHFR IPI00217253 0 2 3 6 0 3 3.3 0.054 Cytoplasm 0.93Ribosomal protein S11 RPS11 IPI00025091 3 5 3 6 0 5 2.7 0.050 Cytoplasm 0.289 xFerritin, heavy polypeptide 1 FTH1 IPI00554521 4 11 0 2 4 7 2.7 0.038 Cytoplasm 0.526 xPoliovirus receptor PVR IPI00219425 2 4 2 3 0 4 2.6 0.093 Cytoplasm 0.754Protein disulfide isomerase family A,member 4

PDIA4 IPI00009904 13 23 12 22 11 51 2.6 0.022 Cytoplasm 0.621 x

Heat shock 60 kDa protein 1 (chaperonin) HSPD1 IPI00472102 8 14 7 18 6 27 2.6 0.014 Cytoplasm 0.331 xDesmoglein 2 DSG2 IPI00028931 6 10 8 11 4 26 2.5 0.062 Plasma membr 0.838* xVacuolar protein sorting 35 homolog(S, cerevisiae)

VPS35 IPI00018931 3 6 4 6 0 5 2.5 0.061 Cytoplasm 0.682 x

Ribonuclease, RNase A family, 4 RNASE4 IPI00029699 0 6 5 12 10 16 2.4 0.035 Cytoplasm 0.543*Glucuronidase, beta GUSB IPI00027745 0 6 2 4 8 12 2.3 0.085 Cytoplasm 0.485 xLeukotriene A4 hydrolase LTA4H IPI00219077 14 26 13 19 7 35 2.3 0.035 Cytoplasm 0.786 xPhosphoribosylglycinamide formyltransferase GART IPI00025273 6 14 7 17 3 5 2.2 0.021 Cytoplasm 0.753* xLectin, galactoside-binding, soluble, 3 LGALS3 IPI00465431 8 21 13 35 18 27 2.2 0.012 Plasma membr 0.678 xSerine/arginine-rich splicing factor 2 SRSF2 IPI00005978 4 6 0 2 3 5 2.1 0.115 Plasma membr 0.375*Suppression of tumorigenicity 14(colon carcinoma)

ST14 IPI00001922 6 9 2 5 4 12 2 0.050 Cytoplasm 0.547

Desmocollin 2 DSC2 IPI00025846 8 13 6 12 5 13 2 0.025 Extracell space 0.77 xHeat shock 70 kDa protein 5(glucose-regulated protein, 78 kDa)

HSPA5 IPI00003362 17 29 10 19 14 34 2 0.008 Plasma membr 0.313* x

87JO

UR

NA

LO

FPR

OT

EO

MIC

S91

(2013)

84–96

88 J O U R N A L O F P R O T E O M I C S 9 1 ( 2 0 1 3 ) 8 4 – 9 6

Interestingly, factors secreted by cancer cells can also play amajor role in the acquisition of drug resistance. For example,interleukin-4 produced by colorectal cancer cells is a majorcytoprotective cytokine that mediates resistance against5-fluorouracil and oxaliplatin [11]. In addition, the culturesupernatant from drug-resistant pancreas carcinoma cellscan protect drug-sensitive cells to chemotherapy [21]. Inbreast cancer, secreted interleukin-6 and -18 contribute todrug resistance [22,23]. Thus, secreted factors in multipletumor types can modulate drug resistance. In addition,analysis of secreted factors also provides insight into intra-cellular processes, as many tumor cells shed intracellular andeven nuclear proteins into the extracellular space [24].

Given the central role that CSCs are thought to play inmediating drug resistance it is important to identify theplethora of factors that are secreted by them (the ‘secretome’).Therefore, we set out to identify secretomedifferences betweenpaired cultures of highly tumorigenic CSCs and their isogenicnon-tumorigenic differentiated progeny. Bioinformatic analysisof the resulting protein lists revealed enrichment of several CSCprotection pathways. In addition, we identified several CSC-secreted drug-modifying enzymes that are shown to modulatedrug responsiveness of nearby non-CSCs.

2. Experimental procedures

2.1. Collection, isolation and expansion of colorectal cancerstem cells and differentiated tumor cells

Collection of tumor specimens was performed as described[12,17]. In short, human colorectal tumor specimens wereobtained from patients undergoing a liver resection formetastatic adenocarcinoma, in accordance with the ethicalcommittee on human experimentation. Informed consentwas obtained from all patients. The obtained tissue fragmentswere washed extensively with phosphate-buffered saline andwere mechanically dissociated using scalpels and vigoroustrituration to yield small fragments (<1 mm3) and single cells.Enzymatic digestion was performed using thermolysin 0.05%(Sigma, Type X) in Dulbecco's modified Eagle's medium/F12containing 5 mm HEPES (Invitrogen, Carlsbad, CA) for 2 h at37 °C. The suspension was then filtered through a 40-μm-poresize nylon cell strainer (BD Falcon) to separate the tissue frag-ments from the single cells. The single cell suspension wascultured in advanced Dulbecco's modified Eagle's medium/F12(Invitrogen) supplemented with 0.6% glucose (BDH Lab.Supplies), 2 mm L-glutamine (Biowhittaker, Rockland, ME),9.6 μg/ml putrescin (Sigma), 6.3 ng/ml progesterone (Sigma),5.2 ng/ml sodium selenite (Sigma), 25 μg/ml insulin (Sigma),100 μg/ml apotransferrin (Sigma), 5 mm HEPES (Invitrogen),0.005 μg/ml trace element A (Cellgro), 0.01 μg/ml traceelement B (Cellgro), 0.01 μg/ml trace element C (Cellgro),100 μmβ-mercaptoethanol (Merck), 10 ml antibiotic–antimycotic(Invitrogen), 4 μg/ml gentamicine (Invitrogen), 0.002% lipid mix-ture (Sigma), 5 μg/ml glutathione (Roche), and 4 μg/ml Heparin(Sigma). Growth factors (20 ng/ml EGF (Invitrogen) and 10 ng/mlb-FGF (Abcam, Cambridge, UK)) were added to the cell culturemedium freshly each week. All cell cultures were carried out innontissue culture treated flasks (BD Falcon) at 37 °C in a 5% CO2

humidified incubator. In vitro differentiation was induced byculturing colon CSC cultures for 3 weeks on collagen-coateddishes in Dulbecco's modified Eagle's medium/F12 (GIBCO)supplemented with 20% fetal bovine serum. Passage numbers ofthe clones used (n = 10) CSC cultures and the accompanyingdifferentiated tumor cell cultures were all below 15.

2.2. Protein extraction from conditioned medium

All differentiated cell cultures were washed three times withserum-free medium to remove all serum residues. Paired CSCand differentiated TC cultures were seeded in 10 cm2 dishesand cultured for 48 h in serum-free Stem Cell Medium(described in [17]). After incubation, the conditioned mediumof both cultures was collected. Floating cells and cellulardebris were removed by centrifugation (2000 rpm, 5 minat 4 °C) and filtration using a 0.22 μm pore size filter. Theresulting conditioned medium was concentrated by ultrafil-tration using “Amicon Ultra-15” (3 kDa cutoff) centrifugalfilter devices according to the manufacturer's instructions(Millipore). Proteinase inhibitors were added to the secretedproteins. Equal amounts of protein (50 μg) were separated onNuPAGE Novex Bis-Tris Mini Gels (Invitrogen). Gels werestained with Coomassie brilliant blue G-250 (Pierce, Rockford,IL) washed and each lane was sliced into ten bands using aband pattern to guide the slicing. The gel slicing and in-geldigestionwere performed in a laminar flow under keratin-freeconditions.

2.3. In-gel digestion

Before MS analysis, separated proteins were in-gel digested asdescribed [12]. Gel lanes corresponding to the different proteinsamples were sliced into ten bands. Cysteine bonds werereduced with 10 mM dithiotreitol for 1 h at 56 °C andalkylated with 50 mM iodoacetamide for 45 min at RT in thedark. Subsequently proteins were digested with 0.06 μg/μltrypsin o/n at 25 °C. Details can be found in [12].

2.4. NanoLC–MS/MS analysis

Peptides were separated by an Ultimate 3000 nanoLC–MS/MSsystem (Dionex LC-Packings, Amsterdam, The Netherlands) asdescribed previously [12]. Eluting peptides were ionized at1.7 kV in a Nanomate Triversa Chip-based nanospray sourceusing a Triversa LC coupler (Advion, Ithaca, NJ). Intact peptidemass spectra and fragmentation spectra were acquired on aLTQ-FT hybrid mass spectrometer (Thermo Fisher, Bremen,Germany). Intact masses weremeasured at resolution 50.000 inthe ICR cell, MS/MS spectra were acquired in the linear ion trapin a top 5 experiment. Instrument settings can be found in [12].

2.5. Database searching, statistics and Ingenuity PathwayAnalysis

MS/MS spectra were searched against the human IPI database3.31 (67511 entries) using Sequest (version 27, rev 12), which ispart of the BioWorks 3.3 data analysis package (ThermoFisher, San Jose, CA). MS/MS spectra were searched with amaximum allowed deviation of 10 ppm for the precursor mass

89J O U R N A L O F P R O T E O M I C S 9 1 ( 2 0 1 3 ) 8 4 – 9 6

and 1 amu for the fragment masses. Methionine oxidation andcysteine carboxamidomethylation were allowed as variablemodifications, two missed cleavages were allowed and theminimum number of tryptic termini was 1. After databasesearching the DTA and OUT files were imported into Scaffold2.01.01 (Proteome software, Portland, OR). Scaffold was used toorganize the gel-band data and to validate peptide identifica-tions using the Peptide Prophet algorithm [25]. Only peptideidentifications with a probability >95% were retained. Subse-quently, the ProteinProphet algorithm [26] was applied andprotein identifications with a probability of >99% with 2peptides or more in at least one of the samples were retained.Proteins that contained similar peptides and could not bedifferentiated based on MS/MS analysis alone were grouped.For each protein identified, the number of spectral counts(identified MS/MS spectra) in each sample (10 gel bands) wasexported to Excel. Spectral counts were normalized by dividingthe spectral counts per protein by the sum of all counts persample andmultiplying by the average sum across all samples.The list of differentially expressed proteins, including foldchanges was imported in the online software package Ingenuity(Ingenuity IPA, version 7.6) and pathway and network analysiswas performed with only direct relationships. Furthermore thelist of differentially expressed proteins was imported in theonline STRING database of known and predicted proteininteractions. The interactions include direct (physical) andindirect (functional) associations between proteins. Confidencelevels were set to high.

2.6. Validation of mass-spectrometry results by Westernblotting

Preparation and collection of conditioned medium of CSC anddifferentiated tumor cell cultures were done as describedabove in Protein extraction from conditioned medium. Lysates ofCSC-cultures and their isogenic differentiated tumor celloffspring were prepared in lysis buffer (20 mM HEPES pH 7.4,1% NP40, 150 mM NaCl, 5 mM MgCl2, 10% glycerol). Equalamounts of protein derived from lysates and conditionedmediumwere run out on NuPAGE Novex Tris-Acetate Mini Gel(Invitrogen) and were analyzed by Western blotting usingantibodies directed against ALDH1A1 (Abcam, Cambridge,UK), BLMH (Sigma-Aldrich, Zwijndrecht, NL) and β-Actin(AC-15, Novus Biologicals, Littleton, CO, USA).

2.7. Maphosphamide and bleomycin cytotoxicity assays

CSC-cultures and their differentiated progeny were culturedin the presence of maphosphamide (Niomech — IIT GmbH,Bielefeld, Germany) or bleomycin (PHARMACHEMIE B, Haar-lem) at the indicated concentrations for 72 h. In addition, forboth maphosphamide and bleomycin experiments, differen-tiated TCs were treated with conditioned medium derivedfrom CSC-cultures. Conditioned medium was harvested fromCSC-cultures after 48 h of culturing. Mitochondrial activitywas evaluated using CellTiter 96® AQueous Non-RadioactiveCell Proliferation Assay (MTS) (Promega). All absorbancevalues are expressed as percentages of vehicle-treated controlwells.

2.8. Flow cytometry and cell sorting

Dead cells were excluded using viability marker 7-aminoactinomycin D (7-AAD) (R&D, Detroit, MI) and celldoublets and clumps were excluded using doublet discrimina-tion gating. Aldefluor®-positive cellswere analyzed according tothe manufacturer's protocol by using Aldefluor® and DEAB(STEMCELL Technologies). The cell sorting experiments wereconducted with FACS Aria IIU (BD), using FACS Diva version 6.1software.

2.9. Statistical analysis

The paired beta-binomial test was used to analyze foldchanges and statistically significant differences betweenspectral count data generated from CSC and their differenti-ated tumor cells. This statistical test takes into account bothtechnical variation and biological variation within a singlestatistical framework [27]. The Student t test (unpaired, 2-tailed)was performed to analyze statistically significant differencesbetween groups (GraphPad, San Diego, CA). Differences with ap-value of less than .05 were considered statistically significant.

3. Results

3.1. Overlap between the CSC secretome and proteome

We have previously reported on the isolation and characteriza-tion of cancer stem cell (CSC)-enriched tumor cell cultures fromhuman colorectal tumors (colonospheres) and the stablydifferentiated cultures that were derived from them [12,17].Three of these paired cultures were used to analyze potentialdifferences in the collection of proteins secreted by CSCs andtheir differentiated offspring respectively. The total of secretedproteins is frequently referred to as the ‘secretome’. Conditionedmedium from all six cell cultures was collected and run out on aSDS-PAGEgel (Fig. 1A).After tryptic digestion the eluted peptideswere analyzed by nanoLC–MS/MS. The total number of uniqueproteins identified inall secretomeswas 1254. Of these, 547wereexpressed in all three pairs and 268 proteins in two of the threepairs (Fig. 1B). Unsupervised clustering of all six cell culturesshows that the isogenic pairs cluster together (Fig. 1C). Super-vised clustering on the basis of proteins enriched inCSC cultures(Table 1, Supplemental Table 1) shows that all CSC cultures anddifferentiated cultures now cluster together (Fig. 1D). Of allproteins identified, 48 proteinswere enrichedmore than 1.5 foldin the CSC cultures of all three tumors (Table 1). An additional100 proteins were enriched more than 1.5-fold in the CSCcultures of two out of three tumors (Supplemental Table 1). Allthe above mentioned proteins enriched in CSC cultures wereanalyzedwith SecretomeP 2.0 as a computational tool to predicttheir secretory potential [28]. This revealed that about 60% of allproteins were potentially secreted by means of the classical ornon-classical secretion pathways (Table 1, Supplemental Table1). In addition, we analyzed our dataset for the presence ofexosomal proteins. Testing our dataset against the HumanExocarta Database 4.1 revealed that 55% of all enriched proteinsin the CSC cultures were predicted to be exosome-derived(Table 1, Supplemental Table 1). Moreover, of all proteins

90 J O U R N A L O F P R O T E O M I C S 9 1 ( 2 0 1 3 ) 8 4 – 9 6

predicted to be secreted by the non-classical pathway 50% werepredicted to be exosome-derived.

Seventy-two percent (903/1254) of the CSC-enrichedsecretome factors were previously identified in the corre-sponding CSC proteomes (Fig. 2A) [12]. Indeed, the percent-ages of proteins derived from the nucleus, the cytoplasm, theplasma membrane and the extracellular space for theseoverlapping proteins were roughly similar in the secretomeand the proteome (Fig. 2B). This suggests that the secretomeprovides an approximate extracellular image of the CSCsthemselves. Three-hundred-fifty-one proteins were identifiedin the secretome alone, but not in the proteome. As may beexpected, extracellular space and plasma membrane proteinswere strongly enriched in this group of secretome-onlyproteins (Fig. 2C).

Proteome

351 903 2016

Secretome

A B Total p

Total s

CExclus

Fig. 2 – Overlap between proteomes and secretomes. (A.) Venn dsecretomes of all three tumors (L145, L146 and L167) including bpercentage of identified proteins in the proteome and secretomecytoplasm, plasma membrane, extracellular space and unknownexclusively detected in the secretome, ordered according to theimembrane, extracellular space and unknown).

3.2. CSCs secrete CSC markers

The list of 148 CSC-enriched secretome proteins (all proteinsdescribed in Table 1, Supplemental Table 1) was analyzed forpotential functional connectivity by Ingenuity PathwayAnalysis(IPA) and extensive PubMed searches (Supplementary Table 2).Strikingly, we found that many putative markers for (cancer-)SCs were identified in the CSC secretomes, including OLFM4,EphB2 and ALDH1A1 (Supplementary Table 2) [29–31].

3.3. The CSC secretome is enriched in survival factors

IPA analysis identified ‘Cancer’ as the most significantlyassociated disease with the CSC secretome. Proteins governingCell Death and Survival regulation were most significantly

Nucleus

Cytoplasm

Plasma Membrane

Extracellular Space

Unknown

23%

59%

9%

3% 6%

19%

50%

15%

12%

4%

roteome

ecretome

11%

31%

23%

29%

6%

ively secretome

iagram of all proteins detected in the proteomes andoth CSC and differentiated cells. (B.) Pie-charts showing the, ordered according to their intracellular localization (nucleus,). (C.) Pie-chart showing the percentage of identified proteinsr intracellular localization (nucleus, cytoplasm, plasma

91J O U R N A L O F P R O T E O M I C S 9 1 ( 2 0 1 3 ) 8 4 – 9 6

enriched in the category ‘Molecular and Cellular Functions’(Supplementary Table 3). Extensive literature analysis revealedthat the vast majority of these (37 out of 43 molecules) wereanti-apoptotic (Supplementary Table 2). This could reflect ageneral increase in CSC survival capacity.

3.4. Protein synthesis, folding, modification, and processing

Supplementary Table 2 shows that many CSC-secretedproteins are involved in protein synthesis, including 10 ribo-somal proteins, 4 tRNA synthetases, a translation initiationfactor (EIF2C2), and a ribosome recycling factor (ABCE1). IPAanalysis further revealed that EIF2 signaling is one of the topcanonical pathways overrepresented in the CSC secretome(Supplementary Table 3). STRING analysis also identified a14-protein network governing protein synthesis (Fig. 3A).

Furthermore, the list of CSC-enriched proteins also con-tains multiple protein chaperones located in the cytosol,in the mitochondria, and in the ER. In the cytosol theseare components of the TCP and the prefoldin complexes

A. Protein synthesis

E. Antioxidan

B. Protein fo

D. Glycolysis

Fig. 3 – STRING analysis of CSC-culture secretomes reveals the prnetworks. All CSC-enriched proteins (148) were analyzed by usinassociation networks. The strength of the associations is represeassociations were set to 0.700 (high). The identified protein netw(C.) protein modification, (D.) glycolysis, and (E.) antioxidant resp

(Supplementary Table 2). The protein folding machinery inthe ER is represented by chaperones and disulfide bridgeisomerases. Some of these factors not only control proteinfolding, but also protect against oxidative ER stress. Indeed,GRP78 and ERp72 play an important role in the unfoldedprotein response (UPR) which is activated upon ER stress.Strikingly, UPR proteins were also found to be upregulated inbreast CSCs [32], suggesting that an increased capacity for ERprotein folding may be an intrinsic property of various typesof CSCs. STRING analysis revealed the presence of twochaperone complexes of 5 and 9 proteins in the CSCsecretome (Fig. 3B).

Both Ingenuity and STRING analyses showed that factorsinvolved in protein modification are also enriched in the CSCsecretome (Supplementary Table 2). These include enzymesinvolved in glycosaminoglycan degradation and glycosylation(Fig. 3C). Although not identified by STRING or Ingenuity as aseparate functional group, we also found 8 proteins thatregulate protein tyrosine phosphorylation in the CSC secretome(Supplementary Table 2).

t respons

lding C. Protein modification

esence of proteomemaintenance, glycolysis, and antioxidantg the STRING web application for functional proteinnted by line thickness. The required confidence scores for allorks control (A.) protein synthesis, (B.) protein folding,onse.

92 J O U R N A L O F P R O T E O M I C S 9 1 ( 2 0 1 3 ) 8 4 – 9 6

3.5. Energy metabolism

IPA analysis also revealed that proteins involved in energymetabolism (Glycolysis/Gluconeogenesis, Pentose phosphateshunt) and glutamate metabolism are enriched in the CSCsecretome. Likewise, STRING analysis identified a functionalprotein network governing glycolysis (Fig. 3D). In general,cancer cells utilize glycolysis rather than oxidative phosphor-ylation to produce ATP, despite sufficient oxygen supply. Thisphenomenon, known as the Warburg effect, may therefore beparticularly relevant to CSCs.

3.6. Oxidative stress

The top Toxicity Pathway identified by IPA is ‘OxidativeStress’ (p = 1.0e − 4) (Supplementary Table 3). All 11 proteinsinvolved in the regulation of oxidative stress are listed inSupplementary Table 2. Interestingly, there is a significantoverlap between the set of proteins in the secretome andthose that are known to be activated by the transcriptionfactor NRF2 (NFE2L2), a key activator of the anti-oxidantresponse (p = 2.7e − 4; CCT7, FTL, GPX2, GCLC, TXNRD1, FTH1)(Supplementary Table 3). This suggests that CSCs may protectthemselves from oxidative stress in part by activating theNRF2 antioxidant pathway. In addition, STRING analysisidentified an antioxidant functional interaction network con-sisting of GPX2, GPX3, GSS, TXNDC12, TXNRD1 and GCLC(Fig. 3E).

3.7. CSCs express ALDH1A1 and BLMH

Extensive PubMed analysis of all secreted factors revealedthat CSCs secrete a number of factors that have been implicatedin the detoxification and/or conversion of different types ofanti-cancer drugs (SupplementaryTable 2). These areALDH1A1,ALDH2, BLMH, CES2, and CD46. Although colonosphere culturesare highly enriched in CSCs, they also contain a subpopulationof differentiated non-tumorigenic cells. By using Aldefluor®, afluorescent probe for aldehyde dehydrogenase activity, thetumorigenic CSCs can be separated from the non-tumorigeniccells by FACS sorting [12,17,33]. FACS sorting based on Aldefluorrevealed that Aldefluorhigh CSCs, but not the non-tumorigenicAldefluorlow cells, selectively express ALDH1A1 and BLHM(Fig. 4A).

3.8. CSCs secrete ALDH activity and protect differentiatedtumor cells from maphosphamide

ALDH1A1 and ALDH2 can detoxify aldehyde-containinganti-cancer drugs such as cyclophosphamide (CPP) andmay play a role in drug resistance [34,35]. Maphosphamideis the active analogue of cyclophosphamide which isfrequently used for in vitro experiments. In line with theresults presented in Fig. 4A we found that ALDH1A1 isexpressed in and secreted by the CSC cultures, but to farlower levels by the isogenic differentiated tumor cellcultures (Fig. 4B). Treatment of two distinct CSC lines(L145; L167) and their differentiated progeny showed thatCSCs were far more resistant to the drug than differentiatedtumor cells (Fig. 4C, D).

To testwhetherALDHactivity secreted fromCSCswas able tocross-protect differentiated tumor cells from maphosphamide,the latter were cultured in the presence of CSC-conditionedmedium, and in the presence or absence of the ALDH-inhibitorDEAB. Maphosphamide treatment (40 μM) killed the vastmajority (72 ± 2%) of differentiated tumor cells. However, CSC-conditioned medium largely neutralized the cytotoxic effect ofmaphosphamide (27 ± 5% cell death; p < 0.05) (Fig. 4E). Additionof DEAB to the conditionedmedium reduced cross-protection bymore than 50%. This suggests that approximately half ofthe protective effect of the CSC-conditioned medium was dueto secreted ALDH activity. Survival-signaling growth factors(Supplementary Table 2) and/or additional drug-detoxifyingenzymes may account for the remaining protective activity inCSC-conditioned medium.

3.9. CSCs secrete bleomycin-hydrolase and protect differ-entiated tumor cells from bleomycin

BLMH (bleomycin hydrolase, a cysteine peptidase) providesprotection against the glycopeptide anti-cancer drug bleomycinby hydrolyzing its first peptide bond [36]. Although bleomycin isnot used for the treatment of CRC, in vitro data have shown thatit is effective particularly against cell lines displaying microsat-ellite instability [37]. Furthermore, the finding that highlevels of BLMH are secreted by CSCs may have implicationsfor the treatment of other types of cancer in whichbleomycin is part of the standard treatment regimen.These include Hodgkin's lymphoma, squamous cell carci-noma and testicular cancer [38].

We first validated that CSCs secrete bleomycin hydrolase(BLMH) to much higher levels than differentiated tumor cells(Figs. 4A and 5A). In line with this, CSCs were markedlyresistant to bleomycin while differentiated tumor cells wereeffectively killed in two distinct cell cultures (L145; L167)(Fig. 5B). Moreover, CSC-conditioned medium protected dif-ferentiated tumor cells from bleomycin (Fig. 5C) as it didagainst maphosphamide (Fig. 4E).

4. Discussion

Normal stem cells usemultiple mechanisms tomaintain theirgenomic integrity over time, without having been exposed tochemotherapy. Recent evidence indicates that intestinalcancer originates from normal tissue SCs [39,40]. By inference,CSCs may be direct (albeit deranged) descendants from tissueSCs and, as such, have inherited the natural resistancepathways that prevent SC exhaustion. The relationshipbetween CSCs and chemoresistance is highly complex andneeds to be investigated for each specific chemotherapeuticdrug. The most straightforward relationship is when CSCsexpress enzymes that directly detoxify anti-tumor drugs, suchas the inactivation of CPP/MPP by CSC-produced ALDHs.However, drug resistance can be acquired in many differentways, for instance through increased expression of drug effluxpumps, antioxidant enzymes and DNA repair enzymes [41].The relationship between CSCs and resistance to a particulardrug will therefore depend on whether or not CSCs expressthe major resistance mechanism to that drug.

diffdiff diffL145 L146 L167 L145 L146 L167

diff diff diffd

ALDH1A1

sph sph sph d

ALDH1A1

prot

eom

eactinactin

C D

d

colonospheres

0 25 32.5 40 500

50

100

maphosphamide (µM) maphosphamide (µM) maphosphamide (µM)

Via

bilit

y (%

of c

ontr

ol)

**

*

0 25 32.50

50

100

**

standard medium

conditioned medium

conditioned medium + DEAB

colonospheres

E

00

25

50

75

100 * *

40

n.s

secr

etom

eA

ALDH1A1

BLMH

L167

low high Aldefluor

B

differentiated TC differentiated TC

sph sph sph

L167L145 L145

Via

bilit

y (%

of c

ontr

ol)

Via

bilit

y (%

of c

ontr

ol)

Fig. 4 – ALDH1A1 is expressed in the tumorigenic Aldefluorhigh fraction of colonosphere cells and secreted ALDH1A1 protectsdifferentiated cells from maphosphamide. (A.) Single cell cultures of L167 colonospheres were separated into Aldefluor®high

and Aldefluor®low cell populations by FACS sorting. Western blot analysis of Aldefluor®high and Aldefluor®low cell populationswas performed to determine ALDH1A1 and BLMH expression in both populations. (B.) Western blot analysis of ALDH1A1 levelsin the secretomes (left) and proteomes (right) of CSC-cultures and differentiated TC (L145, L146 and L167). (C.) and (D.) L145 andL167 CSC-cultures and their differentiated progeny were treated with maphosphamide for 72 h using the indicatedconcentrations. Cell viability was then assessed by MTS assays for mitochondrial activity. Absorbance values (in triplicate) areexpressed as percentage of vehicle-treated control wells. (E.) Differentiated cells (L145) were cultured in normal culturingmedium, conditioned medium derived from CSC-cultures or conditioned medium + DEAB (15 μM) and subsequently treatedwith 40 μM maphosphamide for 72 h. Cell viability was then assessed by MTS assays. Absorbance values (in triplicate) areexpressed as percentage of vehicle-treated control wells. *Statistical significance (unpaired, 2-tailed t test: p < 0.05).

93J O U R N A L O F P R O T E O M I C S 9 1 ( 2 0 1 3 ) 8 4 – 9 6

Our studies have shown that CSCs express and secretehigh levels of many different survival proteins, includingthose that reduce oxidative stress. Recent evidence showsthat breast CSCs are relatively resistant to oxidative stress andthat this prevents their exhaustion [19]. The Nrf2 antioxidantprogram has previously been identified as an importantmediator of chemoresistance in cell lines derived frommultiple tumor types [42–44]. Our results suggest that Nrf2may play a similar role in colon cancer, but also that this mayoccur mainly in the CSC subpopulation. The relatively recentdiscoveries that the Nrf2 program has pro-tumorigenicproperties [45] and that it plays an important role inchemoresistance [42–44] have intensified efforts to searchfor and develop Nrf2 inhibitors [46]. We propose that suchinhibitors may neutralize the intrinsic (Nrf2-mediated) resis-tance of colorectal CSCs to chemotherapy.

Several DNA repair pathways have been implicated in theprotection of stem cells against DNA damage, thereby main-taining genome integrity [47]. Perhaps surprisingly, we have

not found any evidence for the enrichment of DNA repairproteins in the proteome or secretome of colorectal CSCs.Rather, we found many factors governing maintenance ofproteome integrity being enriched in CSCs. These includedfactors involved in protein synthesis, folding, modificationand destruction. An increased capacity for maintaining prote-ome integrity therefore appears to be a novel property of the CSCphenotype. However, we also noted that proteins governingprotein synthesis and foldingare frequently found in secretomes.Therefore, the overrepresentation of this class of proteins in theCSC-secretomes needs to be interpreted with caution and therelevance of this finding for CSCmaintenance needs to be testedexperimentally.

An important new concept from the current study is thatsecretion of specific proteins from CSCs modulates drugresponses in nearby tumor cells. Thus, CSCs not only possessmultiple mechanisms that promote self-preservation, butmay also preserve cells in their direct vicinity by secretingdrug-detoxifying enzymes. Recently, it was shown that normal

secr

etom

e

AL145 L146 L167

BLMH

B

sph diff sph sph

differentiated TC

D

0 25 50 100

colonospheres

Via

bilit

y (%

of c

ontr

ol)

*

L167

0 1000

50

100

bleomycin (µM)

bleomycin (µM) bleomycin (µM)

Via

bilit

y (%

of c

ontr

ol) conditioned medium

standard medium*

L167

0 25 50 100 200 4000

50

100

0

50

100

Via

bilit

y (%

of c

ontr

ol)

L145

** *

colonospheresC

diff diff

differentiated TC

actin

Fig. 5 – Bleomycin hydrolase secreted by CSC-cultures protects differentiated cells from bleomycin. (A.)Western blot analysis ofBLMH expression in the secretomes of L145, L146 and L167 CSC-cultures and their differentiated progeny. (B.) and (C.) L145 andL167 CSC-cultures and their differentiated progeny were treated with bleomycin for 72 h using the indicated concentrations.Cell viability was then assessed by MTS assays. Absorbance values (in triplicate) are expressed as percentage ofvehicle-treated control wells. (D.) Differentiated tumor cells (L167) were cultured in normal culturing medium or conditionedmedium derived from the L167 CSC-culture for 72 h. Cell viability was then assessed by MTS assays. Absorbance values(in triplicate) are expressed as percentage of vehicle-treated control wells. *Statistical significance (unpaired, 2-tailed t test:p < 0.05).

94 J O U R N A L O F P R O T E O M I C S 9 1 ( 2 0 1 3 ) 8 4 – 9 6

intestinal stem cells are flanked by Paneth cells that act ascritical SC support cells [48]. Differentiated tumor cells incolorectal tumorsmay act in the sameway by providing supportto the CSCs from which they originated. Thus, by secretingdrug-detoxifying enzymes CSCsmay provide survival signals totheir own closely associated support cells and this may beessential for their own survival.

To what extent the mechanisms described above areoperating within CRC tumors during chemotherapy needs tobe addressed in future experiments. The colonosphere modelwould be ideal in such studies as they contain CSCs anddifferentiated tumor cells. They are also highly tumorigenic in

mice while preserving the phenotype of the original patienttumor [17].

Supplementary data to this article can be found online athttp://dx.doi.org/10.1016/j.jprot.2013.06.027.

Acknowledgment

The following authors were financially supported by theDutch Cancer Society (KWF): BLE, EJAS, and KMG. WvH wassupported by the Dutch Foundation for Medical ScientificResearch (ZONMW). We acknowledge the VUmc-Cancer Center

95J O U R N A L O F P R O T E O M I C S 9 1 ( 2 0 1 3 ) 8 4 – 9 6

Amsterdam for the financial support of the proteomics infra-structure, TVP and CRJ.

R E F E R E N C E S

[1] Ferlay J, Shin HR, Bray F, Forman D, Mathers C, Parkin DM.Estimates of worldwide burden of cancer in 2008: GLOBOCAN2008. Int J Cancer Dec 15 2010;127(12):2893–917.

[2] Saltz LB, Clarke S, Diaz-Rubio E, ScheithauerW, Figer A, WongR, et al. Bevacizumab in combination with oxaliplatin-basedchemotherapy as first-line therapy in metastatic colorectalcancer: a randomized phase III study. J Clin Oncol Apr 202008;26(12):2013–9.

[3] Hurwitz H, Fehrenbacher L, Novotny W, Cartwright T,Hainsworth J, Heim W, et al. Bevacizumab plus irinotecan,fluorouracil, and leucovorin for metastatic colorectal cancer.N Engl J Med Jun 3 2004;350(23):2335–42.

[4] Broadbridge VT, Karapetis CS, Price TJ. Cetuximab inmetastatic colorectal cancer. Expert Rev Anticancer Ther May2012;12(5):555–65.

[5] Iqbal S, Lenz HJ. Integration of novel agents in the treatmentof colorectal cancer. Cancer Chemother Pharmacol Sep2004;54(Suppl. 1):S32–9.

[6] O'Brien CA, Pollett A, Gallinger S, Dick JE. A humancolon cancer cell capable of initiating tumour growth inimmunodeficient mice. Nature Jan 4 2007;445(7123):106–10.

[7] Ricci-Vitiani L, Lombardi DG, Pilozzi E, Biffoni M, Todaro M,Peschle C, et al. Identification and expansion of humancolon-cancer-initiating cells. Nature Jan 4 2007;445(7123):111–5.

[8] Todaro M, Francipane MG, Medema JP, Stassi G. Colon cancerstem cells: promise of targeted therapy. Gastroenterology Jun2010;138(6):2151–62.

[9] Vermeulen L, Todaro M, de Sousa MF, Sprick MR, Kemper K,Perez AM, et al. Single-cell cloning of colon cancer stem cellsreveals a multi-lineage differentiation capacity. Proc NatlAcad Sci U S A Sep 9 2008;105(36):13427–32.

[10] Pang R, Law WL, Chu AC, Poon JT, Lam CS, Chow AK, et al. Asubpopulation of CD26+ cancer stem cells with metastaticcapacity in human colorectal cancer. Cell Stem Cell Jun 42010;6(6):603–15.

[11] Todaro M, Alea MP, Di Stefano AB, Cammareri P, VermeulenL, Iovino F, et al. Colon cancer stem cells dictate tumorgrowth and resist cell death by production of interleukin-4.Cell Stem Cell Oct 11 2007;1(4):389–402.

[12] Van Houdt WJ, Emmink BL, Pham TV, Piersma SR, VerheemA, Vries RG, et al. Comparative proteomics of colon cancerstem cells and differentiated tumor cells identifies BIRC6 as apotential therapeutic target. Mol Cell Proteomics Dec2011;10(12):M111.

[13] Cho RW, Clarke MF. Recent advances in cancer stem cells.Curr Opin Genet Dev Feb 2008;18(1):48–53.

[14] Dylla SJ, Beviglia L, Park IK, Chartier C, Raval J, Ngan L, et al.Colorectal cancer stem cells are enriched in xenogeneictumors following chemotherapy. PLoS One 2008;3(6):e2428.

[15] Creighton CJ, Li X, Landis M, Dixon JM, Neumeister VM,Sjolund A, et al. Residual breast cancers after conventionaltherapy display mesenchymal as well as tumor-initiatingfeatures. Proc Natl Acad Sci U S A Aug 18 2009;106(33):13820–5.

[16] Li X, Lewis MT, Huang J, Gutierrez C, Osborne CK, Wu MF,et al. Intrinsic resistance of tumorigenic breast cancer cells tochemotherapy. J Natl Cancer Inst May 7 2008;100(9):672–9.

[17] Emmink BL, Van Houdt WJ, Vries RG, Hoogwater FJ, GovaertKM, Verheem A, et al. Differentiated human colorectal cancercells protect tumor-initiating cells from irinotecan.Gastroenterology Jul 2011;141(1):269–78.

[18] Nagaria P, Robert C, Rassool FV. DNA double-strand breakresponse in stem cells: mechanisms to maintain genomicintegrity. Biochim Biophys Acta Feb 2013;1830(2):2345–53.

[19] Diehn M, Cho RW, Lobo NA, Kalisky T, Dorie MJ, Kulp AN,et al. Association of reactive oxygen species levels andradioresistance in cancer stem cells. Nature Apr 92009;458(7239):780–3.

[20] O'Brien CA, Kreso A, Ryan P, Hermans KG, Gibson L, Wang Y,et al. ID1 and ID3 regulate the self-renewal capacity of humancolon cancer-initiating cells through p21. Cancer Cell Jun 122012;21(6):777–92.

[21] Arlt A, Vorndamm J, Muerkoster S, Yu H, Schmidt WE, FolschUR, et al. Autocrine production of interleukin 1beta confersconstitutive nuclear factor kappaB activity andchemoresistance in pancreatic carcinoma cell lines. CancerRes Feb 1 2002;62(3):910–6.

[22] Yao L, Zhang Y, Chen K, Hu X, Xu LX. Discovery of IL-18 as anovel secreted protein contributing to doxorubicin resistanceby comparative secretome analysis of MCF-7 and MCF-7/Dox.PLoS One 2011;6(9):e24684.

[23] Conze D, Weiss L, Regen PS, Bhushan A, Weaver D, Johnson P,et al. Autocrine production of interleukin 6 causes multidrugresistance in breast cancer cells. Cancer Res Dec 152001;61(24):8851–8.

[24] Karagiannis GS, Pavlou MP, Diamandis EP. Cancersecretomics reveal pathophysiological pathways in cancermolecular oncology. Mol Oncol Dec 2010;4(6):496–510.

[25] Shevchenko A, Wilm M, Vorm O, Mann M. Mass spectrometricsequencing of proteins silver-stained polyacrylamide gels. AnalChem Mar 1 1996;68(5):850–8.

[26] Keller A, Nesvizhskii AI, Kolker E, Aebersold R. Empiricalstatistical model to estimate the accuracy of peptideidentifications made by MS/MS and database search. AnalChem Oct 15 2002;74(20):5383–92.

[27] Pham TV, Jimenez CR. An accurate paired sample test forcount data. Bioinformatics Sep 15 2012;28(18):i596–602.

[28] Bendtsen JD, Jensen LJ, Blom N, Von HG, Brunak S.Feature-based prediction of non-classical andleaderless protein secretion. Protein Eng Des Sel Apr2004;17(4):349–56.

[29] van der Flier LG, Haegebarth A, Stange DE, van deWetering M,Clevers H. OLFM4 is a robust marker for stem cells in humanintestine and marks a subset of colorectal cancer cells.Gastroenterology Jul 2009;137(1):15–7.

[30] Merlos-Suarez A, Barriga FM, Jung P, Iglesias M, Cespedes MV,Rossell D, et al. The intestinal stem cell signature identifiescolorectal cancer stem cells and predicts disease relapse. CellStem Cell May 6 2011;8(5):511–24.

[31] Huang EH, Hynes MJ, Zhang T, Ginestier C, Dontu G,Appelman H, et al. Aldehyde dehydrogenase 1 is a marker fornormal and malignant human colonic stem cells (SC) andtracks SC overpopulation during colon tumorigenesis. CancerRes Apr 15 2009;69(8):3382–9.

[32] Bartkowiak K, Effenberger KE, Harder S, Andreas A, Buck F,Peter-Katalinic J, et al. Discovery of a novel unfolded proteinresponse phenotype of cancer stem/progenitor cells from thebone marrow of breast cancer patients. J Proteome Res Jun 42010;9(6):3158–68.

[33] Shenoy A, Butterworth E, Huang EH. ALDH as a marker forenriching tumorigenic human colonic stem cells. MethodsMol Biol 2012;916:373–85.

[34] Emadi A, Jones RJ, Brodsky RA. Cyclophosphamide andcancer: golden anniversary. Nat Rev Clin Oncol Nov2009;6(11):638–47.

[35] Ma I, Allan AL. The role of human aldehyde dehydrogenase innormal and cancer stem cells. Stem Cell Rev Jun 2011;7(2):292–306.

[36] Zheng W, Johnston SA, Joshua-Tor L. The unusual active siteof Gal6/bleomycin hydrolase can act as a carboxypeptidase,

96 J O U R N A L O F P R O T E O M I C S 9 1 ( 2 0 1 3 ) 8 4 – 9 6

aminopeptidase, and peptide ligase. Cell Apr 3 1998;93(1):103–9.

[37] Li HR, Shagisultanova EI, Yamashita K, Piao Z, Perucho M,Malkhosyan SR. Hypersensitivity of tumor cell lines withmicrosatellite instability to DNA double strand breakproducing chemotherapeutic agent bleomycin. Cancer Res Jul15 2004;64(14):4760–7.

[38] Chen J, Stubbe J. Bleomycins: towards better therapeutics. NatRev Cancer Feb 2005;5(2):102–12.

[39] Barker N, Ridgway RA, van Es JH, van de Wetering M, BegthelH, van den Born M, et al. Crypt stem cells as the cells-of-originof intestinal cancer. Nature Jan 29 2009;457(7229):608–11.

[40] Zhu L, Gibson P, Currle DS, Tong Y, Richardson RJ, BayazitovIT, et al. Prominin 1 marks intestinal stem cells that aresusceptible to neoplastic transformation. Nature Jan 292009;457(7229):603–7.

[41] Dean M, Fojo T, Bates S. Tumour stem cells and drugresistance. Nat Rev Cancer Apr 2005;5(4):275–84.

[42] Wang XJ, Sun Z, Villeneuve NF, Zhang S, Zhao F, Li Y, et al. Nrf2enhances resistance of cancer cells to chemotherapeutic drugs,the dark side of Nrf2. Carcinogenesis Jun 2008;29(6):1235–43.

[43] Jiang T, Chen N, Zhao F, Wang XJ, Kong B, Zheng W, et al.High levels of Nrf2 determine chemoresistance in type IIendometrial cancer. Cancer Res Jul 1 2010;70(13):5486–96.

[44] Zhang P, Singh A, Yegnasubramanian S, Esopi D, KombairajuP, Bodas M, et al. Loss of Kelch-like ECH-associated protein 1function in prostate cancer cells causes chemoresistance andradioresistance and promotes tumor growth. Mol CancerTher Feb 2010;9(2):336–46.

[45] DeNicola GM, Karreth FA, Humpton TJ, Gopinathan A, Wei C,Frese K, et al. Oncogene-induced Nrf2 transcription promotesROS detoxification and tumorigenesis. Nature Jul 72011;475(7354):106–9.

[46] Magesh S, Chen Y, Hu L. Small molecule modulators ofKeap1–Nrf2–ARE pathway as potential preventive andtherapeutic agents. Med Res Rev Jul 2012;32(4):687–726.

[47] Tichy ED. Mechanisms maintaining genomic integrity inembryonic stem cells and induced pluripotent stem cells. ExpBiol Med (Maywood) Sep 1 2011;236(9):987–96.

[48] Sato T, van Es JH, Snippert HJ, Stange DE, Vries RG, van denBorn M, et al. Paneth cells constitute the niche for Lgr5 stemcells in intestinal crypts. Nature Jan 20 2011;469(7330):415–8.