Characterization of a Large Panel of Patient-Derived Tumor … · 2012. 9. 16. · Cancer Therapy: Preclinical See commentary by Kopetz et al., p. 5161 Characterization of a Large

Cancer Therapy: PreclinicalSee commentary by Kopetz et al., p. 5161

Characterization of a Large Panel of Patient-Derived TumorXenografts Representing the Clinical Heterogeneity ofHuman Colorectal Cancer

Sylvia Julien1, Ana Merino-Trigo2, Ludovic Lacroix3, Marc Pocard4, Diane Go�er�e3, Pascale Mariani5,Sophie Landron6, Ludovic Bigot3, Fariba Nemati5, Peggy Dartigues3, Louis-Bastien Weiswald7,Denis Lantuas4, Loïc Morgand9, Emmanuel Pham1, Patrick Gonin3, Virginie Dangles-Marie4,7, Bastien Job3,Philippe Dessen3, Alain Bruno6, Alain Pierr�e6, Hugues De Th�e8, Hany Soliman8, Manoel Nunes2, GuillaumeLardier2, Loreley Calvet2, Brigitte Demers2, Gr�egoire Pr�evost1, Patricia Vrignaud2, Sergio Roman-Roman5,Olivier Duchamp9, and Cyril Berthet9

AbstractPurpose: Patient-derived xenograft models are considered to represent the heterogeneity of human

cancers and advanced preclinical models. Our consortium joins efforts to extensively develop and

characterize a new collection of patient-derived colorectal cancer (CRC) models.

Experimental Design: From the 85 unsupervised surgical colorectal samples collection, 54 tumors were

successfully xenografted in immunodeficient mice and rats, representing 35 primary tumors, 5 peritoneal

carcinoses and 14 metastases. Histologic and molecular characterization of patient tumors, first and late

passages on mice includes the sequence of key genes involved in CRC (i.e., APC, KRAS, TP53), aCGH, and

transcriptomic analysis.

Results: This comprehensive characterization shows that our collection recapitulates the clinical

situation about the histopathology and molecular diversity of CRC. Moreover, patient tumors and

corresponding models are clustering together allowing comparison studies between clinical and preclinical

data. Hence, we conducted pharmacologic monotherapy studies with standard of care for CRC (5-fluo-

rouracil, oxaliplatin, irinotecan, and cetuximab). Through this extensive in vivo analysis, we have shown the

loss of human stroma cells after engraftment, observed ametastatic phenotype in somemodels, and finally

compared the molecular profile with the drug sensitivity of each tumor model. Through an experimental

cetuximab phase II trial, we confirmed the key role of KRAS mutation in cetuximab resistance.

Conclusions:This new collection couldbring benefit to evaluate novel targeted therapeutic strategies and

tobetter understand thebasis for sensitivity or resistance of tumors from individual patients.ClinCancer Res;

18(19); 1–15. �2012 AACR.

IntroductionDespite great advances in our understanding of the

molecular basis of colorectal cancer (CRC) and theincreasing number of targeted therapies, the treatmentof advanced CRC frequently remains disappointing, par-

ticularly when facing metastases. This is linked to thenatural history of the tumors and/or patients, considering(i) the interpatient variability in drug exposure, (ii) thediversity of CRC with respect to molecular profile andsensitivity to a specific agent, and (iii) the intratumoralheterogeneity, and the highly variable tumor cell dou-bling time. All these sources of variability make difficultthe validation of markers predictive for the response tochemotherapy.

The interest of patient-derived xenograft (PDX) modelsin preclinical testing is more and more emphasized. Thesexenografts derived directly from patient samples, withoutin vitro manipulation, provide a more accurate depictionof human tumor biologic characteristics than cell line-derived xenografts with hundreds in vitro passages andserial passing across several generations of mice (1).Moreover, as these models might better reflect a clinicalresponse (2–8), several groups have established disease-specific panels of xenografts directly from patient tumors

Authors' Affiliations: 1IPSEN Innovation, LesUlis; 2SanofiVitry-sur-Seine;3Institut de Canc�erologie Gustave Roussy, Villejuif; 4Hopital Lariboisi�ere—AP-HP, INSERM U965, Universit�e Paris-Diderot—Paris 7; 5Institut Curie;6Institut de Recherches Servier, Croissy-sur-Seine; 7IFR71, Universit�eParisDescartes, SorbonneParisCit�e; 8InstitutUniversitaired'H�ematologie,Hopital Saint-Louis, Paris; and 9Oncodesign, Dijon, France

Note: Supplementary data for this article are available at Clinical CancerResearch Online (http://clincancerres.aacrjournals.org/).

Corresponding Author: Cyril Berthet, Oncodesign, 20, rue Jean Mazen,Dijon 21076, France. Phone: 33-380-788-260; Fax: 33-380-788-261;E-mail: [email protected].

doi: 10.1158/1078-0432.CCR-12-0372

�2012 American Association for Cancer Research.

ClinicalCancer

Research

www.aacrjournals.org OF1

Research. on April 10, 2021. © 2012 American Association for Cancerclincancerres.aacrjournals.org Downloaded from

Published OnlineFirst July 23, 2012; DOI: 10.1158/1078-0432.CCR-12-0372

http://clincancerres.aacrjournals.org/

(2, 3, 7–10). The ability to passage patient fresh tumortissues into large numbers of immunodeficient miceprovides possibilities for better preclinical testing of newtherapies.

The above considerations emphasize the need for newin vivo models of CRC representing their genotypic andphenotypic diversities. Here, through a consortium effortfrom hospitals, academic groups, and biotech and phar-maceutical companies, we have developed a large collec-tion of CRC models directly derived from tumor samplescollected during patient surgery. We have extensivelycharacterized these xenograft models at the molecular,histologic, and pharmacologic levels to ensure that theyrepresent the diversity of the clinical situations. All modelcharacteristics and clinical patient history are being com-piled in a web-based database for efficient features searchand interconnection. To assess whether the histopathol-ogy of human CRC is maintained in those models, weinvestigated the level of differentiation, the presence ofmucus and showed the efficient replacement of thehuman stroma by host cells. The high diversity of themolecular profile and pharmacologic response of CRChighlight the difficulties to classify colon tumors uponmolecular profile. Nevertheless, such collection allowedus to set up a phase II trial to investigate the sensitivityprofile of the panel and see whether it correlates withclinical data. Our analysis confirms the link betweencetuximab sensitivity and KRAS wild-type status observedin the clinical setting. This collection could be used toassess new therapeutic approaches, to better understandthe basis for sensitivity of tumors from individualpatients, and potentially help the stratification of patientsaccording to molecular markers.

Materials and MethodsColorectal tumor samples

CRC samples were collected, after patient’s informedconsent, in 3medical centers: Curie Institute (Paris, France),GustaveRoussy Institute (Villejuif, France), andLariboisi�ereHospital (Paris, France). Immediately after surgery (1 hourafter resection in average, except for rectal tumors for whichit can go up to 4 hours), 2 fragments were snap frozenin liquid nitrogen, then stored at �80�C, for molecularcharacterization, 2 fragments were fixed in alcohol-forma-lin-acetic acid solution and paraffin-embedded for histo-pathologic analysis, and 2 fragments were transferred inculture medium including DMEMwith 10mmol/L HEPES,4.5 g/L glucose, 1 mmol/L pyruvate sodium, 200 U/mLpenicillin, 200 mg/mL streptomycin, 200 mg/mL gentami-cin, 5 mg/mL ciprofloxacin, 20 mg/mL metronidazole,25 mg/mL vancomycin, and 2.5 mg/mL fungizone or DMEMwith Nanomycopulitine (Abcys) for engraftment done inthe 24 hours after resection. After engraftment, residualfragments were frozen in DMEM including 10% dimethylsulfoxide and 10% fetal bovine serum. Similar process ofsample conservation was conducted on tumor fragmentscollected frommice and rats. A thawing test from P3 frozensamples was done in a period of 3 months after freezing.

AnimalsAnimals were maintained in the animal facilities of each

institution following standard animal regulation and stricthealth controls allowing transfer between members of theconsortium. Swiss-nude and CB17-SCID female mice, aswell as NIH-nude rats were bred at Charles River France.Mouse weights were more than 18 g and rat weights weremore than 160 g at the start of experiments. Their care andhousingwere in accordancewith institutional guidelines, aswell as national and European laws and regulations as putforth by the French Forest and Agriculture Ministry and thestandards required by the UKCCCR guidelines (11).

Tumor engraftmentAfter 2 to 24 hours following the patient surgery, the

tumor samples were engrafted on 2 Swiss nudemice. Smallfragments (�50 mm3) were subcutaneously engrafted intothe scapular area or on the flank of anesthetized mice(xylazine/ketamine or isoflurane protocol). Tumor growthwasmeasured at least once a week and serial fragment graftsof each given tumor were conducted on 3 to 5 Swiss nude orCB17-SCID (after 3 passages)micewhen the tumors reach avolume of 800 to 1500 mm3. Engraftment on nude rat wasconducted for each tumor after 3 successfully passages onnudemice. To increase tumor take rate, nude rats were wholebody irradiated at 7Gywith a g-source, 24 to48hoursbeforetumor grafting. For orthotopic implantation, analgesia wasinduced 30minutes before implantation by intraperitoneal(i.p.) injection of Flunixin (5mg/kg bodyweight; Finadyne,Schering-Plough). Then, mice were anesthetized by i.p.injection of xylazine (5 mg/kg body weight; Rompun,Bayer) and Tiletamine (30 mg/kg body weight; Zoletil,

Translational RelevanceWith the development of targeted therapies, the need

is growing for preclinical models characterized at themolecular level and allowing correlation with drug sen-sitivity and patient clinical history. The lack of surgerysamples available for research, the small size of biopsiesand the multiplicity of resource centers are a limitationfor standardized sample preparation required to identifycomplex prognostic signatures. To develop our collec-tion of patient-derived colorectal tumor models, weorganized a global process frommultiple clinical centersto pharmaceutical companies. We showed the mainte-nance of the clinical features, even with the highlyheterogeneous colorectal cancers. Using our panel ofmodels, we reproduced the results observed in clinicaltrials designed for cetuximab, with respect to KRASstatus. Moreover, we observed the occurrence of metas-tases that emphasizes the use of patient-derived xeno-graft models as a powerful investigational platform toevaluate new therapies and generate samples in thesearch for molecular signature.

Julien et al.

Clin Cancer Res; 18(19) October 1, 2012 Clinical Cancer ResearchOF2




Virbac). For implantation, tumor fragment was attached tothe serosa of the cecum to be entirely surrounded by theserosa.

Sample preparationFrozen fragments were cut in a cryostat at �20�C

then beginning and end sections were stained with Hae-matoxylin-Eosin-Saffron (HES) for histologic control andevaluation of tumor cell percentage by pathologists. Onlysamples with an average percentage above 40% were usedfor DNA extraction. Genomic DNAwas extracted accordingto QIAamp DNA mini-Kit protocols (Qiagen). Total RNAwas extracted from tumor samples using a TRIzol method(Invitrogen), purified with an RNeasy mini kit (Qiagen).

Histologic characterizationSerial 5-mm-thick sections were cut from each paraffin-

embedded tumor. Examination of tumor morphology wasconducted on slices processed for HES staining. Tumordifferentiation, necrosis, stroma, and presence of mucuswere scored for each xenograft and correspondent patient’stumor biopsy. Eight to ten different samples were reviewedfor each xenograftmodel at 2different passages (betweenP6andP12). Aperio Scan ScopeXT scannerwas used to acquirewhole sections images at�200magnification. Alu probe insitu hybridization was processed. On 5-mm-thick sections ofparaffin-embedded tumors in a Discovery XT biomarkerplatform (Roche Diagnostics) according to the manufac-turer’s specifications and using its reagents, including theAlu Probe SN59560. Briefly, sections are deparaffinated,rehydrated, pretreated by protease digestion (20 min at37�C), denaturated 8 min at 85�C, incubated with the Aluprobe 1 hour at 50�C, washed, postfixed, stained by animmunoenzymaticmethodusing alkaline phosphatase andNBT/BCIP (nitro-blue-tetrazolium/5-bromo-4-chloro-3-indolyl-phosphate) as a substrate, and counterstained withNuclear Fast Red.

Molecular characterizationMicrosatellite Sequence Instability status. Microsatellite

sequence instability (MSI) testing was conducted accordingto the National Cancer Institute guidelines. A 5-microsat-ellite consensus panel was used as (12, 13).Gene sequencing. On the basis of the literature, we

selected exons 2 to 11 of TP53; exons 9 and 16 of APC;exon 3 of AKT1, exons 11 and 15 of BRAF, exons 4 to 11 ofFBXW7, exon 3 of CTNNB1, exons 2 and 3 of KRAS, exons10 and 21 of PIK3CA, exons 18 to 21 of EGFR, exons 2, 26–27 and 30 of KDR, exon 4 of FCGR2A and FCGR3A, andexon 3 of ERCC1 for direct sequencing analysis, conductedafter PCR amplification (Supplementary Table S1). PurifiedPCR products were sequenced using BigDye TerminatorCycle Sequencing Kit (Applied Biosystems). Sequencingreactionswere analyzedon48-capillary 3730DNAAnalyzerin both sense and antisense directions. All mutationsdetected were controlled with independent amplificationat least once. Sequences reading and alignment were con-ducted with SeqScape software (Applied Biosystems).

Oligonucleotide aCGH and gene expression analysis.Genomic DNA was analyzed using the Human GenomeCGHMicroarray-244A (Agilent Technologies). In all experi-ments, sex-matched DNA from a pooled human female ormale (Promega) was used as a reference. OligonucleotideaCGH processing was conducted as detailed in the manu-facturer’s protocol (version 4.0; http://www.agilent.com).Data were extracted from scanned images using the FeatureExtraction software (v10.2.1.2 to 10.7.3.1, Agilent), alongwith protocols CGH_v4_10_Apr08 to CGH_105_Jan09.Acquired signals were normalized according to their dyeand local GC% content using in-house scripts under the Rstatistical environment (http://cran.r-project.org). Result-ing log2(ratio) were segmented using the circular binarysegmentation (14) algorithm implementation from theDNA copy package for R. Aberration status calling wasautomatically conducted for each profile according to itsinternal noise [variation of log2(ratio) values across con-secutive probes on the genome]. All genomic coordinateswere established on the UCSC human genome build hg19(15). To compensate the possible variation of dynamics inthe profiles of paired samples, a simple linear regression fitwas conducted under R, to increase the lower dynamicsprofile to the higher dynamics one. Hierarchical clusteringwas conducted under R on segmented data, using Euclideandistances and Ward’s construction method. The geneexpression analysis was conducted using a GeneChipExpression 30-Amplification Reagents Kit and U133A Gen-eChip arrays (Affymetrix). All data were imported intoResolver software (Rosetta Biosoftware) for database man-agement and quality controls. Moreover, the raw data (.CELfiles) were imported in the BrB Arrays Tools (http://linus.nci.nih.gov/BRB-ArrayTools.html) and normalized with aRobustMulti-array Average (RMA) procedure (http://www.bioconductor.org) to conduct the class prediction analysesand functional analysis. The microarray data related to thisarticle have been submitted to the Array Express data repos-itory at the European Bioinformatics Institute (http://www.ebi.ac.uk/arrayexpress/) under the accession number E-MTAB-991 forGeneExpression andE-MTAB-992 for aCGH.

In vivo pharmacologic studiesDrugs. 5-fluorouracil (5-FU) (ICN) and oxaliplatin

(Sanofi) were diluted in a 5% (w/v) solution of glucose inwater. Cetuximab (Imclone) was diluted in a phosphatebuffer solution. Irinotecan (Pfizer) was diluted in water.

Chemotherapy. Protocol design, chemotherapy techni-ques, and methods of data analysis were essentially equiv-alent to those described previously (16). On the first day oftreatment, the animals bearing 100 to 200 mm3 tumorswere unselectively distributed to the various treatment andcontrol groups (n ¼ 8–10 per group).

Oxaliplatin and 5-FU were IV administered at a dosagecorresponding to 70% of the highest nontoxic dose(HNTD) in mice (50 and 5 mg/kg/injection, respectively)with a Q4Dx2 schedule of administration (i.e., 2 injectionswith a 4-day interval). Irinotecan (CPT-11) was IV injectedat a dose of 22mg/kg/injection, Q2Dx3 (70%ofHNTD). In

Colorectal Patient-Derived Tumor Models

www.aacrjournals.org Clin Cancer Res; 18(19) October 1, 2012 OF3




Table 1. Clinical characteristics and in vivo patient-derived tumor take rate

Primary tumor samples (n ¼ 58)

Patient clinical characteristics

All collectedGrowing on nudemice (n ¼ 35)

Parameters Class NumberFrequency(%) Number

Frequency(%)

In vivo tumortake rate (%)

GenderFemale 36 62.1 22 62.9 61.1Male 22 37.9 13 37.1 59.1

Age<50 y 8 13.8 7 20.0 87.5>50 and <60 y 14 24.1 8 22.9 57.1>60 and <70 y 13 22.4 5 14.3 38.5>70 and <80 y 17 29.3 10 28.6 58.8>80 y 6 10.3 5 14.3 83.3

pT - primary tumor statusUnknown 0 0 0 0 NApTx 0 0 0 0 NApTis 2 3.4 0 0 0pT1 1 1.7 0 0 0pT2 8 13.8 3 8.6 37.5pT3 36 62.1 25 71.4 69.4pT4 11 19.0 7 20.0 63.6

pN - lymp node statusUnknown 0 0 0 0 NApNx (not assessable) 0 0 0 0 NApN0 27 46.5 9 25.7 33.3pNþ 31 53.5 26 74.3 83.9

pM - distant metastasis statusUnknown 0 0 0 0 NApMx (not assessable) 2 3.4 1 2.9 50.0pM0 35 60.3 20 57.1 57.1pM1 21 36.2 14 40.0 66.7

StageUnknown 2 3.5 1 2.9 50.00 2 3.5 0 0 0I 7 12.1 2 5.7 28.6II 12 20.7 5 14.3 41.7III 14 24.1 13 37.1 92.9IV 21 36.2 14 40.0 66.7

CEA dosage before surgery (ng/mL)Unknown 19 32.8 12 34.3 63.2<6 21 36.2 9 25.7 42.9>6 18 31.0 14 40.0 77.8

Primary tumor locationProximal colon 20 34.5 12 34.3 60.0Distal colon 20 34.5 14 40.0 70.0Rectum 18 31.0 9 25.7 50.0

(Continued on the following page)

Julien et al.





Table 1. Clinical characteristics and in vivo patient-derived tumor take rate (Cont'd )

Metastasis and carcinosis tumor samples (n ¼ 27)

Patient clinical characteristics

All collected Growing on nudemice (n ¼ 19)

Parameters Class Number Frequency(%)

Number Frequency(%)

In vivo tumortake rate (%)

GenderFemale 16 59.3 11 57.9 68.7Male 11 40.7 8 42.1 72.7

Age<50 y 3 11.1 1 5.3 33.3>50 and <60 y 8 29.6 5 26.3 62.5>60 and <70 y 9 33.3 7 36.8 77.8>70 and <80 y 5 18.5 4 21.0 80.0>80 y 2 7.4 2 10.5 100

pT - primary tumor statusUnknown 2 7.4 0 0.0 0.0pTx 1 3.7 1 5.3 100.0pTis 0 0.0 0 0.0 NApT1 0 0.0 0 0.0 NApT2 2 7.4 2 10.5 100.0pT3 10 37.0 8 42.1 80.0pT4 12 44.4 8 42.1 66.7

pN - lymp node statusUnknown 2 7.4 0 0.0 0.0pNx (not assessable) 4 14.8 3 15.8 75.0pN0 6 22.2 4 21.0 66.7pNþ 15 55.6 12 63.2 80.0

pM - distant metastasis statusUnknown 2 7.4 0 0.0 0.0pMx (not assessable) 0 0.0 0 0.0 NApM0 2 7.4 2 10.5 100.0pM1 23 85.2 17 89.5 73;9

StageUnknown 2 7.4 0 0.0 0.00 0 0.0 0 0.0 NAI 0 0.0 0 0.0 NAII 2 7.4 2 10.5 100.0III 0 0.0 0 0.0 NAIV 23 85.2 17 89.5 73.9

CEA dosage before surgery (ng/mL)Unknown 9 33.3 9 47.4 100.0<6 3 11.1 2 10.5 66.7>6 15 55.6 8 42.1 53.3

Primary tumor locationProximal colon 9 33.3 6 31.6 66.7Distal colon 16 59.3 12 63.2 75.0Rectum 2 7.4 1 5.3 50.0

(Continued on the following page)






nude rats, 5-FU, oxaliplatin, irinotecan were injected IVat the HNTD (5-FU at 30–40 mg/kg/injection, Q7Dx3,oxaliplatin at 4 mg/kg/injection, Q4Dx3, and irinotecan at40 mg/kg/injection, Q7Dx3). Cetuximab was administeredat the dose of 40 mg/kg/injection (IP, Q4Dx4) in mice and10 mg/kg/injection (IV Q7Dx3) in rats. Tumors were mea-sured with a caliper 2 to 5 times weekly. Tumor volumeswere estimated from 2-dimensional tumor measurements:Tumor volume (mm3)¼ [length (mm)�width (mm)2]/2.For ethical reasons the animal bearing an excessive tu-mor volume (>2,000 mm3 for mice and 10,000 mm3 forrats) or tumors impinging too much their behavior wereeuthanized.

Tumor growth inhibition (DT/DC value). Becausetumors were measurable at the start of therapy, the initialtumor burden was taken into account in the calculation ofthe tumor growth inhibition (DT/DC value): DT/DC (%) ¼[(median TDayY �median TDayX)/(median CDayY �medianCDayX)] � 100 (where DayY is the day of evaluation, andDayX is the day of initiation of therapy for treated [T] andcontrol [C] tumor volumes).

Tumor regressions. Tumor regressions can be either par-tial (more than 50% reduction in tumor volume) or com-plete (tumor regression below the limit of palpation).

Determination of the tumor Doubling Time. The tumordoubling time (in days; Td) was estimated from the plot ofthe log linear growth of the control group tumors inexponential growth (100–1,000 mm3 range).

Statistical analysisRecursive partitioning method was conducted using SAS

JMP v9 software as described in ref. 17. The Fisher’s test, c2

test, and all logRank analyses were conducted using EverstatV5 (Sanofi based on SAS 8; SAS Institute Inc.).

ResultsClinical characteristics of the patients and tumor takein mice

A total of 85 CRC samples from primary tumors, perito-neal carcinoses, or distant metastases were collected fromunsupervised patients and subcutaneously engrafted intonude mice. Patient clinical characteristics are presentedin Table 1. The PDX collection nicely reflects the diversityof colorectal clinical cases (18) in terms of gender, age,carcinoembryonic antigen (CEA) dosage, primary tumorlocation, lymph node, and distant metastasis status asdetermined at primary stage. It has to be noted that thelowest stage of the disease, namely, stage 0 (e.g., in situ), isnot present within the collection.

Of the 85 engraftments, 54 led to the establishment ofPDX models, i.e., a tumor take rate of 63.5%. We noted ahigher tumor take rate for metastases than primarytumors, albeit it was not statistically significant. Noimpact of the cold ischemic time (less than 24 hours)was noticed and the median time to reach 300 mm3 afterthe first engraftment was 59.4 days (see Supplementary

Table 1. Clinical characteristics and in vivo patient-derived tumor take rate (Cont'd )

Parameters associated within vivo tumor take rate No growth Growth

ProbabilitiespNþ and CEA � 6 and proximalor distal colon

0.0000 1.0000

pNþ and CEA � 6 and rectum 0.1250 0.8750pNþ and CEA < 6 and pM0 0.1667 0.8333pN0 and rectum 0.8750 0.1250

NumberspNþ and CEA � 6 and proximalor distal colon

0 21

pNþ and CEA � 6 and rectum 1 7pNþ and CEA < 6 and pM0 1 5pN0 and rectum 7 1

NOTE: Patient clinical characteristics and associated in vivo tumor take rate of (i) all collected samples and (ii) xenograft modelsestablished on nude mice for primary tumor samples (upper table) and metastasis and carcinosis tumor samples (middle table).Metastasis metastatic samples include hepatic, splenic, and mesenteric lymph node metastasis.Abbreviations: pTx, primary tumor cannot be assesse;. pTis, primary in situ tumor; pT1, tumor invades mucosa and submucosa; pT2,tumor invadesmuscularis propria; pT3, tumor invades serosa, subserosa, or pericolic fat tissues; pT4, tumor invades peritoneal cavitythroughserosaor expands toother proximal organs throughserosa; pN0,N0, nopositive regional lymphnodes; pNþ, at least 1 positiveregional lymph node; pM0, at the stage of primary tumor evaluation, no distant metastasis; pM1, at the stage of primary tumorevaluation, presence of distant metastasis.Influencing parameter associations significantly correlated with tumor take rate as determined by recursive partitioning (bottom table).The probabilities of no growth/growth and the effective for each signature are presented.

Julien et al.





Table S2). A biobank was established for all models andthe thawing success rate reaches 92.3%, close to ourtumor take rate on fresh samples, 97.2% at P3 to P6.Among this collection, 5 coupled primary tumor/associ-ated synchronous hepatic metastasis, 1 coupled perito-neal carcinoses/splenic metastasis, and 1 triplet primarytumor/peritoneal carcinoses and mesenteric lymph nodemetastasis derived from the same patient have beenestablished. Overall, the collection contains 35 primarytumor, 5 peritoneal carcinoses, and 14 metastatic (12hepatic, 1 splenic, and 1 mesenteric lymph node) models.Recursive partitioning was applied to identify parameters

associatedwith high or low probability of in vivo tumor takerate (Table 1). None of the prior treatments impact signif-icantly the tumor take rate. Two or three combined para-meters were associated with higher probability of in vivotumor take rate: (i) pNþ, CEA � 6 ng/mL (probability ¼ 1while combined with primary tumor location ¼ proximalor distal colon, n¼ 21; probability¼ 0.87 while combined

with primary location¼ rectum, n¼ 8) and (ii) pNþ, CEA <6 ng/mL, pM0 (probability ¼ 0.83, n ¼ 6). Conversely, thesignature pN0þ primary tumor location ¼ rectum is asso-ciated with a low probability of in vivo tumor take rate(probability ¼ 0.12, n ¼ 8), which could be impacted to alonger hot ischemic time.

Histopathologic diversity of the colorectal tumorcollection

Histopathologic characterization of our panel showed aconcordance between xenografts (2 different passagesbetween P6 and P12) and the corresponding patient’stumor in termof tumor differentiation andmucus secretion(Fig. 1A and Fig. S1). Using standard clinical parameters,xenografts were classified in well to moderately differenti-ated (92%) and moderately differentiated to undifferenti-ated (8%). Ten percent of the xenografts were classified asmucinous adenocarcinomas, a raremorphologic subtype ofcolorectal adenocarcinoma in which more than 50% of the

Figure 1. Preservation of the tumorphenotype, histopathologicdiversity, and stromacharacterization of the colorectaltumor collection. A, comparison ofthe 3 main histologic colorectaladenocarcinoma subtypes observedbetween the patient tumor and itscorresponding derived xenograft atthe indicated passage. Originalmagnification: �200. B, loss of thehumanstroma in xenografts is shownin tumor CR-IGR-011C with ALUprobe in situ hybridization. a, tumorof the patient: stromal and tumor cellnuclei are positive (dark blue). b,xenograft (eighth passage): stromalcells are negative (nuclei unstained),tumor cells are positive. Arrow:stroma; arrowhead: tumor cells;scale bar: 50 mm.

Patient: CR-IGR-0034PPatient: CR-LRB-0010P Patient: CR-LRB-0008M

P10: CR-LRB-0010P P11: CR-LRB-0008M P11: CR-IGR-0034P

Pat

ient

Tum

or m

odel

Well to moderate Well differentiateddifferentiated

Poor to undifferentiatedA

Tumor of the patient Xenograft (eighth passage)

a b

B






Analysis done MSS Microsatellite stability No changes in gene copiesNo data available MSI-L Low microsatellite instability Gain of 1 (or more) copies

MSI-H High microsatellite instable Loss of 1 (or less) copiesMutation Wild type Drug response scoring criteria:

- = ΔT/ΔC > 42 %Model from + = 10 % < ΔT/ΔC 42 %the same patient ++ = -10 % ΔT/ΔC 10 % (tumor stabilization)

+++ = ΔT/ΔC -10 % or PR, CR observed (tumor regression)

Mo

del

co

de

TP

53P

53 fo

nct

ion

nal

test

AP

CK

RA

SP

IK3C

AF

BX

W7

BR

AF

CT

NN

B1

EG

FR

AK

T1

MS

I sta

tus

Ear

ly p

assa

ge

Lat

e p

assa

ge

TP

53

AP

C

KR

AS

PIK

3CA

FB

XW

7

BR

AF

CT

NN

B1

EG

FR

AK

T1

PT

EN

ME

T

5-F

U

l-OH

P

CP

T-1

1

CE

TU

XIM

AB

Tu

mo

r

Ear

ly p

assa

ge

Lat

e p

assa

ge

CR-IC-0002P 0 1 1 1 0 1 0 0 0 0 0 MSS 1 0 -1 0 0 0 0 1 -1 1 0 0 1 + - +++ +++ 0 1 1

CR-IC-0003P 0 1 nt 1 0 0 0 0 0 0 0 MSS 1 1 -1 -1 -1 0 -1 0 0 0 -1 -1 0 ++ - - +++ 1 0 1

CR-IC-0004M 2 1 1 1 1 0 0 0 0 0 0 MSS 1 1 1 0 1 1 -1 1 1 1 0 0 1 - - + ++ 1 1 1

CR-IC-0005P 0 0 nt 0 0 0 0 0 0 0 0 MSS 0 0 - - nt + 0 0 0

CR-IC-0006M 0 1 1 0 0 0 0 0 0 0 0 MSI-L 1 1 -1 0 0 1 0 1 0 1 0 0 1 ++ - +++ +++ 0 1 0

CR-IC-0007M 2 1 nt 1 1 0 0 0 0 0 0 MSS 1 1 -1 -1 0 -1 -1 0 -1 1 -1 -1 0 - - ++ +++ 1 0 1

CR-IC-0008P 0 0 nt 1 0 0 0 0 0 0 0 MSS 1 1 -1 -1 1 -1 0 1 -1 1 -1 -1 1 - + +++ - 1 0 0

CR-IC-0009M 0 0 nt 1 0 0 0 0 0 0 0 MSS 1 1 0 -1 1 0 0 1 0 1 0 -1 1 + - + - 1 0 1

CR-IC-0010P 0 0 nt 1 0 0 0 0 0 0 0 MSS 1 1 -1 0 1 0 -1 1 0 1 -1 0 1 - - ++ - 1 0 0

CR-IC-0013M 0 0 0 1 1 0 0 1 0 0 0 MSI-L 1 1 0 0 1 0 0 0 0 0 0 0 0 + + +++ +++ 1 1 1

CR-IC-0016M 2 1 nt 0 0 1 0 0 0 0 0 MSS 1 1 -1 -1 -1 1 -1 1 1 1 -1 0 1 + - +++ + 1 0 1

CR-IC-0018P 0 1 1 1 1 0 0 0 0 0 0 MSS 0 0 0 0 0 0 0 0 0 0 0 0 0 nt nt nt nt 0 1 0

CR-IC-0019P 0 1 1 0 0 0 0 0 0 0 0 MSI-L 1 1 -1 -1 -1 0 0 1 0 1 -1 -1 1 + - +++ +++ 1 1 0

CR-IC-0020P 0 1 1 1 0 0 1 0 0 0 0 MSS 1 0 -1 -1 -1 0 -1 1 0 1 -1 0 1 + - ++ +++ 1 1 0

CR-IC-0021M 0 1 1 1 0 0 1 0 0 0 0 MSS 0 1 - + + +++ 0 1 0

CR-IC-0022P 1 1 nt 1 0 0 1 0 0 0 0 MSS 0 0 + - +++ - 0 0 1

CR-IC-0025M 0 1 1 1 0 0 0 0 0 0 0 MSS 0 0 + - +++ +++ 0 1 0

CR-IC-0028M 0 1 1 0 0 0 0 0 0 0 0 MSS 1 1 - - ++ - 1 1 0

CR-IC-0029P 0 1 1 0 0 0 0 0 0 0 0 MSS 1 1 -1 -1 -1 -1 0 0 -1 0 -1 -1 0 + - +++ - 0 1 0

CR-IC-0032P 0 1 nt 1 0 0 1 0 0 0 0 + - +++ +++ 0 0 0

CR-IGR-0002C 0 1 nt 0 0 0 0 0 0 0 0 MSI-L 1 1 -1 -1 -1 0 -1 0 0 0 -1 0 0 ++ - ++ ++ 0 1 1

CR-IGR-0002P 0 1 nt 0 0 0 0 0 0 0 0 MSI-L 1 1 -1 -1 -1 1 -1 0 1 0 -1 0 0 + - +++ ++ 1 0 1

CR-IGR-0003P 0 1 nt 0 1 0 0 0 0 0 0 MSS 1 1 -1 -1 1 0 -1 0 -1 1 -1 -1 0 + - ++ +++ 1 0 1

CR-IGR-0007P 2 1 1 1 1 0 0 0 0 0 0 MSS 1 1 -1 -1 0 0 -1 1 -1 1 1 0 1 - - + + 0 1 1

CR-IGR-0008P 0 1 nt 1 0 0 0 0 0 0 0 MSI-L 1 0 -1 -1 1 -1 0 0 0 0 1 -1 0 + - nt - 0 0 0

CR-IGR-0009P 2 1 nt 0 1 0 0 0 0 0 0 MSI-L 1 1 -1 0 0 0 0 0 0 0 0 0 0 + + + - 1 0 1

CR-IGR-0011C 2 0 0 0 1 0 0 0 0 0 0 MSS 1 1 0 0 1 0 0 0 0 0 1 0 0 - - +++ - 1 1 1

CR-IGR-0012P 0 0 nt 1 1 0 1 0 0 0 0 MSS 1 1 -1 0 0 0 0 0 -1 0 0 0 0 + - - - 1 0 1

CR-IGR-0014P 1 0 0 1 1 0 0 0 0 0 0 MSS 1 1 1 0 0 0 0 0 0 0 0 -1 0 ++ - - - 0 1 1

CR-IGR-0016P 2 1 nt 0 1 0 0 0 0 0 0 MSS 1 1 -1 0 0 1 -1 1 -1 1 1 0 1 ++ - +++ - 0 0 1

CR-IGR-0023M 2 1 1 1 1 1 0 0 0 0 0 MSS 1 1 -1 0 -1 0 -1 1 -1 1 -1 0 1 + - + + 0 1 1

CR-IGR-0025P 0 1 1 1 1 0 0 0 0 0 0 MSS 0 1 - - + - 0 1 0

CR-IGR-0029P 2 0 0 1 1 1 0 0 0 0 0 MSS 1 1 -1 0 0 0 -1 0 -1 0 1 0 0 + - + - 1 1 0

CR-IGR-0032P 0 0 0 1 1 1 1 0 0 0 0 MSS 1 1 0 -1 0 -1 -1 1 -1 1 -1 0 1 ++ + ++ - 0 1 1

CR-IGR-0034P 3 1 1 0 0 0 0 1 0 0 0 MSS 1 1 -1 -1 0 0 -1 -1 -1 1 0 0 -1 - - + - 1 1 1

CR-IGR-0038C 2 1 1 0 1 0 0 0 0 0 0 MSS - nt nt ++ 1 1 0

CR-IGR-0039P 2 0 0 0 1 0 0 0 0 0 0 MSS nt nt nt nt 0 0 0

CR-IGR-0043P 0 0 0 1 1 0 0 0 0 0 0 MSS 1 1 -1 1 -1 0 0 1 0 1 -1 0 1 - - +++ - 0 1 0

CR-IGR-0047P 0 1 1 0 0 0 0 0 0 0 0 MSS 1 1 -1 1 -1 0 -1 1 0 1 -1 0 1 + - + +++ 0 1 1

CR-IGR-0048M 0 1 1 0 0 0 0 0 0 0 0 MSS 1 1 -1 1 0 0 0 1 0 1 -1 1 1 + - +++ + 0 1 1

CR-IGR-0052M 2 1 1 0 0 0 0 0 0 0 0 MSI-L 1 1 -1 -1 -1 0 -1 0 -1 0 -1 -1 0 + - ++ + 0 1 0

CR-LRB-0003P 0 0 0 1 0 0 0 0 0 1 0 MSI-H 1 1 0 0 0 0 0 0 0 0 1 0 0 + - + - 1 1 0

CR-LRB-0004P 0 0 0 1 0 0 0 1 0 0 0 MSI-H 1 1 0 0 0 0 0 0 0 0 0 0 0 ++ - + - 1 0 1

CR-LRB-0007P 0 1 1 1 0 0 0 0 0 0 0 MSI-L 1 1 -1 -1 1 1 -1 1 -1 1 -1 -1 1 + - + - 0 1 1

CR-LRB-0008M 2 0 0 0 1 1 0 0 1 0 0 MSS 1 0 0 0 0 0 0 0 0 0 0 0 0 + - - + 0 1 1

CR-LRB-0009C 2 0 0 0 1 1 0 0 1 0 0 MSS 1 1 0 0 0 0 0 0 0 0 0 0 0 + - + - 0 1 1

CR-LRB-0010P 0 1 1 1 1 0 0 0 0 0 0 MSS 1 1 -1 0 0 0 -1 0 -1 0 -1 -1 0 - - ++ - 1 1 1

CR-LRB-0011M 0 1 1 1 1 0 0 0 0 0 0 MSS 1 1 -1 1 1 1 -1 1 -1 1 -1 0 1 + - +++ - 0 1 1

CR-LRB-0013P 0 1 1 1 1 0 0 0 0 0 0 MSS + - +++ - 1 1 0

CR-LRB-0014P 0 1 1 1 1 0 0 0 0 0 0 MSS 1 1 0 0 1 0 -1 1 0 1 0 0 1 + - - +++ 1 1 1

CR-LRB-0017P 0 1 1 1 1 0 0 0 0 0 0 MSS 1 1 1 1 1 0 -1 1 -1 1 -1 1 1 + - + - 0 1 0

CR-LRB-0018P 0 1 1 0 0 0 0 1 0 0 0 MSI-H 1 0 0 -1 0 0 0 1 0 0 1 0 1 - - +++ - 0 1 0

CR-LRB-0019C 2 1 1 1 0 0 0 0 0 0 0 MSS 1 1 -1 -1 0 1 -1 1 0 1 -1 0 1 - - ++ +++ 0 1 0

CR-LRB-0022P 0 1 1 0 1 0 0 0 0 0 0 MSI-H - - +++ - 1 1 0

Mutation status CGH array analysis Pharmacology µA data

Pat

ien

t tre

atm

ent

Figure 2. Extensivemolecular and pharmacology analysis of the 54 patient-derived tumormodels. Sequencing,MSI status, and aCGHdatawere generated onpatient tumors or passage P0/P1, and on late passage (aCGH) following the protocol and quality control described in Material and Methods. Pharmacologicstudies were conducted between passages 7 to 9. Tumor-bearing CB-17 SCID mice were treated as described and tumor growth inhibition was evaluatedaccording toNCI standards, aDT/DC�42%being theminimal level to declare antitumor activity (�¼ inactive >42%,þ¼ active�42%,þþ¼�10% <DT/DC� 10%, corresponding to a tumor stabilization, þþþ ¼��10% and/or partial or complete regressions, corresponding to cytoreductive antitumor activity).

Julien et al.





tumor volume is composedofmucin. These resultsmatchedthe classification of their corresponding patient tumors. Wedetermined as well the percentage of stroma and showedthat the tumor cells are clearly enriched in xenograftscompared with patient samples (Supplementary Fig. S1).This enrichment has been observed after the first passage(data not shown), but once established, a constant ratio ofstroma in the tumor among the different passages wasmaintained. Finally, in situ hybridization with Alu probespecific to human cells showed the complete loss of humanstroma in early passages (Fig. 1B). Overall, these observa-tions confirm that our collection recapitulates the histologicprofile and the diversity of CRC.

Molecular characterization of tumors and xenograftmodelsTo investigate molecular abnormalities displayed by our

PDX models and their potential variation occurring duringxenograft take and passages, we determined MSI status andconductedmutational analysis, aCGH, and gene expressionprofiling.When the tumor cellularity of patient sampleswasinsufficient, the molecular characterization was conductedon the corresponding first passage xenograft (P0).

Direct sequencingof genes frequentlymutated in coloncancerSequence of exons described in Material and Methods

was analyzed in all established xenografts localized athotspots and described in the COSMIC Database(ref. 19; Fig. 2).TP53was found tobemutated in38 samples(70%). We detected 25 different mutations, most of themreported as deleterious in IARCdatabase (www-p53.iarc.fr).Full concordance was found when TP53 status was deter-mined by functional analysis of separated alleles in yeast(FASAY, described in ref. 20). About APC, we identified 30differentmutations in 32mutated samples (59%) including29mutations resulting in a truncated protein. For KRAS, wefound 9 different mutations in 26 mutated samples (48%)corresponding to frequent hotspotmutations in codons 12,13, and 61. Two infrequent mutations in codons 19 and 20(L19F and T20S) were found in a single sample, CR-IC-0018P. In the case of PIK3CA, classical mutations E545KandH1047R, in exons 10 and 21 respectively, were detectedin 7 samples (13%). Six samples harbored FBXW7 muta-tions corresponding to 3 missense mutations and 2 muta-tions resulting in a truncated protein. A mutation of BRAFwas observed in 4models (7%) corresponding to 3 hotspotV600E mutations and 1 infrequent mutation (D594N).Other observed gene mutations include infrequent dele-tions ormissensemutations in the b-catenin encoding gene(CTNNB1) and EGFR (D761N), respectively, and unknownmutations in ERCC1 or FCGR2A (data not shown).Overall, frequency of mutations in our collection is

similar to the frequency reported in the literature for coloncancer (Table 2). Identical mutation profile was found inxenografts derived from different tumor sites (primarytumor and carcinosis, synchronic, or metachronic metas-tasis) of the same patient

Characterization of chromosomal abnormalities byhigh-density aCGH

The 244K-oligonucleotide arrays were used to analyzecopy number alterations. The analyzed samples displayedmore than 50% of tumor cells and consisted of 43 earlypassages (P for patient sample, P0 or P1) and 39 latepassages (P6–P9) xenografts including 38 early/late pairedsamples. As depicted in Fig. 3A, all the samples displayedabnormal aCGH profiles with highly recurrent gains andlosses. These recurrent chromosomal abnormalitiesinclude a whole loss of chromosomes 4, 14, 15, 21, and22; a whole gain (more than 3 copies) of chromosomes 7,13, and 20; and a loss of the long arm and gain of shortarm in chromosomes 8 and 17 (Supplementary Fig. S2).About the MSI, 13/53 samples (24.5%) were MSI-Low(MSI-L) or MSI-High (MSI-H) and no correlation wasfound between MSI status and number of alteration onaCGH or percentage of genome with abnormalities (datanot shown).

We examined the status of chromosomal regions corre-sponding to genes classically associated to colon tumori-genesis (Fig. 2). Although single copy modifications werefrequent at these regions, nohigh amplification, or total losswere observed. EGFR, MET, and BRAF were each gained inmore than about a half of samples (56%, 51%, and 51%,respectively), and rarely lost (2% or less of samples). KRASand PIK3CA were gained in 28% and 19% of samples, andlost in identical proportion (29% and 19%, respectively).TP53, FBXW7, AKT1, APC,CTNNB1, and PTENwere lost in70%, 53%, 51%, 47%, 40%, and 29% of samples, respec-tively and very rarely gained.

To examine the genetic stability of the xenograftsthrough passages, we conducted unsupervised clusteringbased on aCGH results obtained with the 38 pairedsamples. Only 4 paired samples did not cluster together(Supplementary Fig. S3). Among them, 2 samples dis-played a very low number of abnormalities that couldresult in misclassification. The 2 other samples displayed

Table 2. Comparison of mutation frequency insamples collected to generate xenograft modeland frequency in scientific literature (on thebasisof the same exon analysis from Cosmicdatabase)

Gene Observed (n ¼ 54) Cosmic

N (%) Data

TP53 38 (70%) 43%APC 32 (59%) 34%KRAS 26 (48%) 34%PIK3CA 7 (13%) 11%FBXW7 6 (11%) 9%BRAF 4 (7%) 12%CTNNB1 2 (4%) 5%EGFR 1 (2%) 1%AKT1 0 (0%) <1%






very high number of abnormalities and could be consid-ered as highly unstable. We also directly compared pro-files of early and late passages after dynamic adjustment(Fig. 3A). Most of the models displayed very similarprofile with some dynamic modification for abnormali-ties. Nevertheless, some new abnormalities could beidentified in late passages and could be attributed toeither a clonal enrichment or to de novo acquisition.Overall, our data show that CRC xenografts display, in

general, a high genomic stability through at least the 10first passages.

Gene expression profilingTo further characterize our models, we conducted gene

expression analysis with Affymetrix U133A-microarrays on115 samples corresponding to 40 patient tumors, 39 earlypassage xenografts (P0 or P1, 37 models) and 36 latepassages xenografts (P7–P9, 29 models). Unsupervised

CR-IGR-0011C (P1 vs. P8)A

B

CR-IC-0019P (P0 vs. P6)

CR-LRB-0014P (P0 vs. P7)

CR-IC-0010P (P0 vs. P8)

Tumor typesourcepassage

Figure 3. Molecular profile comparison of primary tumor and xenograft at early and late passages. Comparison of aCGHand unsupervised clustering basedongene expression of patient tumors and patient-derived tumor xenograft profiles. A, for each couple of xenograft, aCGHprofile havebeen dynamically adjusted(as described in Material and Methods) to be comparable. X axis correspond to the position on genome from chromosome 1 to chromosome X. Y axiscorrespond to log2 ratio, positive values correspond to gain of copies and negative values correspond to loss of copies. Red areas correspond to genomefraction with abnormalities significantly different between early (green curve) and late passage (blue curve). Two first samples (CR-IGR-0011C and CR-IC-0019P) correspond to very stable models whereas other 2 examples (CR-IC-0010P and CR-LRB-0014P) correspond to models with some modification ingenome abnormalities through passages. B, unsupervised clustering based on gene expression of primary tumor and xenograft. This clustering showsdistinct groupof samples. Annotation under clustering tree describes sample name, tumor type (P¼primary tumor;M¼metastasis; C¼carcinosis), source ofsamples (P¼ patient tumor; X¼ xenograft), and passages (T¼ patient tumor; F¼ early or first passage; L¼ late passage). All of patient tumors are clusteredtogether whereas xenograft samples show good correlation between each early-late passage couples.

Julien et al.





clustering based on Pearson correlation with gene filteredon P value over 10�3 clearly separates patient tumors fromxenografts but no major differences between early and latepassage were observed (Fig. 3B). Xenografts from the samemodel clustered together as for aCGH. To identify the set ofgenes that discriminate patient tumors and xenografts, weconducted class comparison between tumors and earlypassage xenografts. Analysis revealed an 800 probe set(614 genes) more than 22,000 on array, differentiallyexpressed with a P value under 10�3 and with fold changemore than 2. Gene ontology analysis of gene differentiallyexpressed was conducted on 203 annotated gene ontologyfunctions. Among the genes downregulated in xenograftswith respect to patient’s tumors, we observed, as expected,enrichment in genes encoding for extracellular matrix com-ponents, collagens, and immune system regulators (Sup-plementary Table S4). In parallel with the lack of humanstroma observed by immunohistology, these results strong-ly suggest that modification of gene expression is to belargely attributed to the loss of the human stromal compo-nents occurring during engrafting and tumor homing ofhost cells such as immunity and stromal cells. Interestingly,thedifferences in gene expression through the passageswererelatively low, consistent with a remarkable stable geneexpression profile of these models.

Heterogeneity of response to pharmacologic agentsTo identify the intrinsic chemosensitivity of each tumor

model to single agent treatments, CB17-SCIDmice bearingestablished PDX were treated with either cytotoxic chemo-therapy (5-FU, oxaliplatin, or irinotecan), or with cetux-imab (Fig. 2).With doubling times ranging from2.8 to 17.0days (median: 6.0 days), the tumor growth rate of thexenografts was slower than most of cell line-derived xeno-grafts. The cachexia induced by the PDX was variable andwas not related to tumor growth kinetics, with a medianbody weight loss of about 8% (range ¼ 0–19%).As a single agent, 5-FU inhibited the tumor growth in

35/52 tumor models. Nevertheless, taking into accountclinical endpoints, such as tumor stabilization (score þþ)or tumor regressions (score þþþ), only 7 models werefound sensitive to 5-FU, with 10% to 40%partial regressionbeing observed in 2 models. The antitumor activity ofoxaliplatin was modest in this panel of colon tumors, withDT/DC values ranged from 12% to 38% in only 5 tumormodels. Conversely, irinotecan showed a significantactivity in 90% of the models (44/49) regardless the originof the tumor. Tumor stabilization was observed in 20%(10/49) whereas tumor regression was achieved in 39%(19/49) of the models. Finally, 25/52 tumor models (48%)responded to cetuximab: tumor stabilization was achievedin 4 tumor models, and tumor regressions in 14 tumormodels, long-term tumor free survivors being observed in5/14 xenografts.The chemosensitivity of these tumor models was also

evaluated in nude rats. In the 44 established models,the tumor growth rate in nude rats was similar to that ofsevere combined immunodeficient (SCID) mice. Although

we studied the chemosensitivity of only 4 nude rat models,we have observed similar ranges of tumor responses inmiceand rats with the 3 cytotoxic agents (data not shown). Forcetuximab sensitivity, only 1 model (CR-LBR-0022P) wasevaluated in nude rat and was moderately sensitive (DT/DC¼ 19%) whereas CR-LRB-0022Pmodel was not sensitive tocetuximab in SCID mice (DT/DC ¼ 78%). Taking together,these results show the heterogeneity in tumor response tochemotherapy and the lack of simple clustering in regards toclinical, histologic, or molecular profiles.

Response to cetuximab versus mutational statusAs illustrated in Fig. 4A, cetuximab was shown very active

against wild-typeKRAS tumors (e.g., CR-IC-0002P), but alsoonmutatedKRASones (e.g., CR-IC-0013M).We investigatedthe relationship between molecular profile and cetuximabsensitivity (Fig. 4B). The survival curves, corresponding to thetime to reach 750 mm3, are presented for cetuximab-treatedversusnontreatedmice inwild-typeKRASxenografts (Fig.4B-1), for cetuximab-treated versus nontreated mice in mutantKRAS xenografts (Fig. 4B-2). A significant difference wasobserved in the 2 cases using a logRank test for factor groupfor parameter J750 for each model; follow by a Chi2 globaltest (P < 0.0001; P < 0.0001). For cetuximab-treated micebearing wild-type KRAS xenografts versus mutant KRASxenografts (Fig. 4B-3), a significant difference was observedusing a logRank test (P < 0.0001).

Nevertheless, about 42% (12/28) of the PDX displayingwild-type KRAS tumors were not responsive to cetuximab.As previously suggested by Sartore-Bianchi and colleagues(21), alterations in other genes of epidermal growthfactor receptor (EGFR) pathway could also explain theresistance to anti-EGFR therapy. As shown in Table 3, theabsence of response to cetuximab was significantly cor-related with the mutational status taking altogether themutations in BRAF, PIK3CA, and KRAS genes (P ¼ 0.02,Fisher’s test). We can also outline that among the tumorsfor which no mutation in KRAS, BRAF, and PIK3CA havebeen found, there are 5 tumors harboring a loss of at least1 copy of PTEN (Fig. 2) suggesting that other genes areresponsible for cetuximab resistance observed.

Metastasis development from orthotopic engraftmentOf 41 models fully analyzed following engraftment into

the cecum of SCID mice, 13 (32%) gave rise to metastases(at least 1 metastatic foci) mainly in mesenteric lymphnodes, liver, and lung (see Fig. 4C and Supplementary TableS3). Metastases appeared at various time, mainly between 2and 3 months following engraftment. From the panel, noclear correlation appears between the nature of the samples(primary tumor vs. metastasis) and their ability to metas-tasize. It is noteworthy that the secondary sites ofmetastaticdissemination of theses colorectal tumors are the same asthose commonly seen in the clinic (i.e., lymph nodes andlivermainly).Hence, someorthotopic PDXcouldbeused asspontaneously metastasis models, which is of great impor-tance for the evaluation of antimetastatic compounds.Further characterizations are ongoing to evaluate the






A

0

200

400

600

800

1,000

1,200

1,400

10 30 50 70

Days postimplantation

Tum

or v

olum

e (m

m3 )

Tum

or v

olum

e (m

m3 )

CR-IC-0002P

0

200

400

600

800

1,000

1,200

1,400

10 30 50 70 90

Days postimplantation

CR-IC-0013M

B

C

1 2

3Hepatic metastases

Lymph node metastases(mesenteric)

Logrank test for factor GROUP

Logrank test for factor GENOTYPE

0 25 50 75J750

100 125 150

0 25 50 75J750

100 125 150 175

0 25 50 75J750

100 125 150 175

STRATA: GROUP=CetuximabGROUP=Ctrl

STRATA: GROUP=CetuximabGROUP=Ctrl

Censored GROUP=CetuximabCensored GROUP=Ctrl

STRATA: GENOTYPE=MutGENOTYPE=Wt

Censored GENOTYPE=MutCensored GENOTYPE=Wt

Censored GROUP=CetuximabCensored GROUP=Ctrl

1.00

0.75

0.50

0.25

0.00Sur

viva

l dis

trib

utio

n fu

nctio

n

1.00

0.75

0.50

0.25

0.00Sur

viva

l dis

trib

utio

n fu

nctio

n

1.00

0.75

0.50

0.25

0.00Sur

viva

l dis

trib

utio

n fu

nctio

nLogrank test for factor GROUP

Figure 4. Cetuximab sensitivity of patient-derived colon tumor xenografts, molecular correlation, and survival analysis in respect of the KRAS status.Metastases development after orthotopic engraftment. A, cetuximab activity observed in CR-IC-0002P (wild-type KRAS) and CR-IC-0013M (mutatedKRAS) models. Black curves correspond to control mice and red curves to treated ones. Red arrows indicate days of treatment. B, survival analysis andKRASmutation status. 1, LogRank analysis onKRASwild-type populations treated or notwith cetuximab. 2, LogRank analysis onKRASmutated populationstreated or not with cetuximab (treated vs. Ctrl). 3, LogRank analysis on KRAS mutated and wild-type populations all treated with cetuximab. Ctrl: control.C, liver and lymph node metastases derived from tumors engrafted into the cecum. Pictures after necropsy (left panels) and after H&E staining ofhistologic section (right).

Julien et al.





metastatic penetrance and potential discrimination withnonmetastatic PDX models.

DiscussionAppropriate preclinical experimental models are neces-

sary to evaluate the efficacy of new anticancer drugs, ther-apeutic combinations, and to identify biomarkers forsensitivity. The human colorectal cancer cell line–derivedxenograft models have been shown to fail in adequatelypredicting clinical response both in disease and compound-oriented settings (22). To implement predictive CRC ani-mal models, we have generated, through a consortium, alarge panel of tumor xenografts established by directlygrafting patient tumor fragments into immunodeficientmice. Our panel includes 54 different models and consti-tutes the most well-annotated panel of colorectal PDX.This panel is representative of the heterogeneity of humancolorectal cancer in terms of clinical parameters, histopa-thology, molecular pattern, and sensitivity to approveddrugs. Importantly, some of the models retain the abilityto develop spontaneousmetastases that is highly needed forpreclinical evaluation of new compounds.Colorectal tumors display a good tumor take rate in

immunodeficient mice (more than 60%) as compared withbreast cancers (10–37%; refs. 7, 10) or prostate cancers (lessthan 5%; ref. 23). As described for other xenograft series (9,10), tumor stage seems to play a major role in tumor takerate. Therefore, node infiltration, advanced stage, and ele-vatedCEA in serumpositively correlate with tumor take rateinmice. In addition, nomajor biaswith respect to colorectalcancer subgroups (as defined by localization, histologic, ormolecular parameters) was introduced by grafting. This isnot the case in breast cancer in which triple negative tumorsare selected in the grafting process (10). Importantly, theproportion of xenografts exhibiting: (i) some of the mostlydescribed somatic mutations (APC, KRAS, BRAF, TP53,FBXW7, or PIK3CA) and (ii) low or high microsatelliteinstability is roughly similar to the proportion found inhuman CRC.As already described for other PDX (4, 6–10), the histo-

logic pattern of our panel is well preserved when compared

with patient tumors and is consistent with the clinicaldiversity of human colorectal pathology. The conservationof the architecture of original tumors, including the lym-phatic and blood vasculature (unpublished data), is a keyadvantage of these models with respect to the cell line–derived xenografts. For instance, only 8% of PDX weremoderately differentiated to poorly differentiated and10%mucinous adenocarcinomas, which is consistent withthe low prevalence of these histologic subtypes. It has beenshown that the stroma component of PDX becomesmainlymurine after a few passages (8). We confirmed this obser-vation, nonetheless the tumor cell enrichment, strikingsimilarity of tumor xenograft architecture, and originaltumors suggest that cancer cells could educate the micro-environment by reprogramming murine stroma cells totheir benefit.

High-density aCGH and gene expression profile micro-arrays data allowed us to address whether significant differ-ences exist between patient tumors and their respectivexenografts. Unsupervised clustering based on aCGH datashowed that most of the xenograft samples clusterized withtheir patient counterpart. Although the genomic alterationsare similar, we could detect in the xenografts alterationsabsent in primary tumors. This being already described inother xenograft series (24, 25), one can speculate thatminorcell populations of the primary tumors might be amplifiedin the xenografts by the grafting process, or alternatively thatthemutational process continued to evolve in the xenograft.It has been recently reported that a triple negative breastcancer xenograft retained all primary tumor mutations anddisplayed amutation enrichment pattern that paralleled thepatient metastasis (26).

Gene expression profile displayed by primary tumors andxenografts were very similar. The set of genes differentiallyexpressed reflect most probably the absence of humanstroma in the xenografts. It would be very interesting toexamine the gene expressionprofile of themurine stromabyhybridizing the xenograft samples inmouse arrays. In breastcancer xenografts, a partial recapitulation of stroma-relatedgene expression bymurine stromahas been found (25). Thestability of xenograft models with passages is crucial forestablishing a preclinical platform. When early and rela-tively late passages of xenografts were compared, no majordifferences could be observed in terms of copy numberalterations or gene expression profile. This clearly suggeststhe stability of colorectal xenograft models, as depicted forbreast and pancreatic tumor xenografts. Therefore, thesemodels could be used for drug candidate evaluation.

Pharmacologic studies conducted with drugs currentlyused in CRC allowed us to validate the value of our xeno-graft panel for evaluating novel drugs. Single agents wereused and revealed a high degree of pharmacologic hetero-geneity of the panel. Overall, the best responses were foundwith irinotecanwith tumor regressions in almost 40%of themodels. As a single agent in patients with newly diagnosedCRC, irinotecan has been reported to generate responserates in the range of 19% to 32% (27). Treatment withcetuximab resulted in tumor regressions in 23% of the

Table 3. Correlation between cetuximabsensitivity and mutation profile of genesinvolved in the EGFR/KRAS pathway

Cetuximab activity score � þ þþ þþþMUTATIONS KRAS 11 1 2 3

KRAS/PI3KCA 3 2 0 0KRAS/BRAF 0 0 0 1BRAF 3 0 0 0PI3KCA 0 1 0 0

EGFR mutated pathway 17 4 2 4EGFR wild-type pathway 10 3 2 10






models. In light of the clinical data, better responses in thewild type KRAS tumors (48%) were expected. It is interest-ing to note that a similar proportion (42.4%) of KRASwild-type cetuximab nonresponsive tumors has been found byBertotti and colleagues in their collection ofmetastatic CRC(7). The comparison of the survival curves of cetuximab-treated mice (Fig. 4B) shows a better survival for wild-typeKRAS versus mutated KRAS tumors was observed. Interest-ingly, a positive correlation was observed between the lackof response to cetuximab and the mutational status KRAS,BRAF, and PIK3CA genes.

In our goal to develop a platform approach, these tumorswere also successfully engrafted on nude rat, known toexhibit metabolism and drug pharmacokinetic/pharmaco-dynamic profiles closer to human ones than mice (28).Although nude rats are less immunodeficient than nudemice, take rate, tumor growth, and pharmacologic profileswere comparable. Therefore, these PDX could be used in therat for drug efficacy studies and pharmacokinetic/pharma-codynamic studies.

Pancreatic PDX have been reported to constitute a veryvaluable pharmacologic tool for drug development, inparticular through a remarkable correlation between drugactivity in xenografts and clinical outcome, both in terms ofresistance and sensitivity (6). Moreover, such PDX modelshave been used to identify CRC patient subpopulation thatcould benefit fromHER2-targeted therapy (7). All together,these results support the use of well-characterized PDXmodels as a powerful investigational platform for efficienttherapeutic decision-making steps in the clinic throughbiomarkers identification. This panel is currently being used

in a prospective way for target expression and evaluation ofpersonalized therapies by the members of the consortium.

Disclosure of Potential Conflicts of InterestFor commercial purposes, an exclusive license for the patient-derived

tumor xenografts described in this article has been granted by the CReMECconsortium to Oncodesign. No potential conflicts of interest were disclosedby the other authors.

AcknowledgmentsThe authors gratefully acknowledge the patients who accepted to con-

tribute to this research program and their family. We also thank the biologicresource centers and departments of histopathology of Curie Institute,Gustave Roussy Institute, and Lariboisi�ere Hospital for their most importantcontribution in the patient tumor collection. The authors are thankful toJean-Jacques Fontaine, Sophie Chateau-Joubert, and Jean-Luc Servely fortheir support in ALU hybridization experiment. The authors are grateful to

Canc�eropole Ile-de-France and Medicen Paris-Region biocluster for provid-ing support in the consortium coordination. The authors really appreciatedthe help of Emmanuel Canet, Antoine Bril, Philippe Genne, Gilbert Lenoir,Christine Perret, Sylvie Robine, Christophe Thurieau, and Gilles Vassal fordiscussions, advice, and continuous support. The authors are thankful to PhilKasprzyk and Francis Bichat for comments on the manuscript.

Grant SupportAll the authors are members of the CReMEC (Center of Resource for

Experimental Models of Cancer) consortium, which has benefited from afinancial support of the FrenchMinistry of Industry. This grantwas part of the"Fonds unique interminist�eriel" program (FUI) and initially selected byMedicen-Paris Region biocluster. In addition, the work conducted at theGustave Roussy Institute was supported by "D�epartement du Val de Marne."

The costs of publication of this article were defrayed in part by thepayment of page charges. This article must therefore be hereby markedadvertisement in accordance with 18 U.S.C. Section 1734 solely to indicatethis fact.

Received February 3, 2012; revised June 18, 2012; accepted July 12, 2012;published OnlineFirst July 23, 2012.

References1. Cree IA, Glaysher S, Harvey AL. Efficacy of anti-cancer agents in cell

lines versus human primary tumour tissue. Curr Opin Pharmacol2010;10:375–9.

2. Fichtner I, Slisow W, Gill J, Becker M, Elbe B, Hillebrand T, et al.Anticancer drug response and expression of molecular markers inearly-passage xenotransplanted colon carcinomas. Eur J Cancer2004;40:298–307.

3. Fiebig HH, Maier A, Burger AM. Clonogenic assay with establishedhuman tumour xenografts: correlation of in vitro to in vivo activity as abasis for anticancer drug discovery. Eur J Cancer 2004;40:802–20.

4. Hidalgo M, Bruckheimer E, Rajeshkumar NV, Garrido-Laguna I, DeOliveira E, Rubio-Viqueira B, et al. A pilot clinical study of treatmentguided by personalized tumorgrafts in patients with advanced cancer.Mol Cancer Ther 2011;10:1311–6.

5. Krumbach R, Sch€uler J, Hofmann M, Giesemann T, Fiebig H-H,Beckers T. Primary resistance to cetuximab in a panel of patient-derived tumour xenograft models: activation of MET as one mecha-nism for drug resistance. Eur J Cancer 2011;47:1231–43.

6. Rubio-Viqueira B, Hidalgo M. Direct in vivo xenograft tumor model forpredicting chemotherapeutic drug response in cancer patients. ClinPharmacol Ther 2008;85:217–21.

7. Bertotti A, Migliardi G, Galimi F, Sassi F, Torti D, Isella C, et al. Amolecularly annotated platform of patient-derived xenografts ('xeno-patients') identifies HER2 as an effective therapeutic target in cetux-imab-resistant colorectal cancer. Cancer Discov 2011;1:508–23.

8. DeRose YS, Wang G, Lin YC, Bernard PS, Buys SS, Ebbert MT, et al.Tumor grafts derived from women with breast cancer authentically

reflect tumor pathology, growth, metastasis and disease outcomes.Nat Med 2011;17:1514–20.

9. N�emati F, Sastre-Garau X, Laurent C, Couturier J, Mariani P, Desjar-dins L, et al. Establishment and characterization of a panel of humanuveal melanoma xenografts derived from primary and/or metastatictumors. Clin Cancer Res 2010;16:2352–62.

10. Marangoni E, Vincent-Salomon A, Auger N, Degeorges A, Assayag F,de Cremoux P, et al. A new model of patient tumor-derived breastcancer xenografts for preclinical assays. Clin Cancer Res 2007;13:3989–98.

11. Workman P, Aboagye EO, Balkwill F, Balmain A, Bruder G, Chaplin DJ,et al. Guidelines for the welfare and use of animals in cancer research.Br J Cancer 2010;102:1555–77.

12. Suraweera N, Duval A, Reperant M, Vaury C, Furlan D, Leroy K, et al.Evaluation of tumor microsatellite instability using five quasimono-morphic mononucleotide repeats and pentaplex PCR. Gastroenterol-ogy 2002;123:1804–11.

13. Umar A, BolandCR, Terdiman JP, Syngal S, de laChapelle A, R€uschoffJ, et al. Revised Bethesda Guidelines for hereditary nonpolyposiscolorectal cancer (Lynch syndrome) and microsatellite instability.J Natl Cancer Inst 2004;96:261–8.

14. Olshen AB, Venkatraman ES, Lucito R, Wigler M. Circular binarysegmentation for the analysis of array-based DNA copy number data.Biostatistics 2004;5:557–72.

15. Karolchik D, BaertschR,DiekhansM, Furey TS,HinrichsA, LuYT, et al.The UCSC genome browser database. Nucleic Acids Res 2003;31:51–4.

Julien et al.





16. Corbett TH, Leopold WR, Dykes DJ, Roberts BJ, Griswold DP Jr,Schabel FM Jr. Toxicity and anticancer activity of a new triazineantifolate (NSC 127755). Cancer Res 1982;42:1707–15.

17. Breiman L, Friedman JH, Olshen RA, Stone CJ. Classification andRegression Trees. New York: Chapman and Hall; 1984.

18. Boyle P, Leon ME. Epidemiology of colorectal cancer. Br Med Bull2002;64:1–25.

19. Forbes SA, BhamraG, Bamford S, Dawson E, Kok C, Clements J, et al.The catalogue of somatic mutations in cancer (COSMIC). Curr ProtocHum Genet 2008;Chapter 10:Unit 10.11.

20. Flaman JM, Frebourg T, Moreau V, Charbonnier F, Martin C, ChappuisP, et al. A simple p53 functional assay for screening cell lines, blood,and tumors. Proc Natl Acad Sci U S A 1995;92:3963–7.

21. Sartore-Bianchi A, Martini M, Molinari F, Veronese S, Nichelatti M,Artale S, et al. PIK3CA mutations in colorectal cancer are associatedwith clinical resistance to EGFR-targeted monoclonal antibodies.Cancer Res 2009;69:1851–7.

22. Voskoglou-Nomikos T, Pater JL, Seymour L. Clinical predictive valueof the in vitro cell line, human xenograft, andmouse allograft preclinicalcancer models. Clin Cancer Res 2003;9:4227–39.

23. Van Weerden WM, Bangma C, De Wit R. Human xenograft models asuseful tools to assess the potential of novel therapeutics in prostatecancer. Br J Cancer 2009;100:13–8.

24. Bergamaschi A, Hjortland GO, Triulzi T, Sørlie T, Johnsen H, ReeAH, et al. Molecular profiling and characterization of luminal-like andbasal-like in vivo breast cancer xenograft models. Mol Oncol2009;3:469–82.

25. Reyal F,GuyaderC,DecraeneC, LucchesiC, AugerN,AssayagF, et al.Molecular profiling of patient-derived breast cancer xenografts. BreastCancer Res 2012;14:R11.

26. Ding L, Ellis MJ, Li S, Larson DE, Chen K, Wallis JW, et al. Genomeremodelling in a basal-like breast cancer metastasis and xenograft.Nature 2010;464:999–1005.

27. Rothenberg ML. Efficacy and toxicity of irinotecan in patients withcolorectal cancer. Semin Oncol 1998;25:39–46.

28. O'Reilly T, McSheehy PM, Kawai R, Kretz O, McMahon L, Brueggen J,et al. Comparative pharmacokinetics of RAD001 (everolimus) in normaland tumor-bearing rodents. Cancer Chemother Pharmacol 2010;65:625–39.






Published OnlineFirst July 23, 2012.Clin Cancer Res Sylvia Julien, Ana Merino-Trigo, Ludovic Lacroix, et al. Human Colorectal CancerXenografts Representing the Clinical Heterogeneity of Characterization of a Large Panel of Patient-Derived Tumor

Updated version

10.1158/1078-0432.CCR-12-0372doi:

Access the most recent version of this article at:

Material

Supplementary

http://clincancerres.aacrjournals.org/content/suppl/2012/09/27/1078-0432.CCR-12-0372.DC2

http://clincancerres.aacrjournals.org/content/suppl/2012/07/23/1078-0432.CCR-12-0372.DC1Access the most recent supplemental material at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected] at

To order reprints of this article or to subscribe to the journal, contact the AACR Publications

Permissions

Rightslink site. (CCC)Click on "Request Permissions" which will take you to the Copyright Clearance Center's

.http://clincancerres.aacrjournals.org/content/early/2012/09/16/1078-0432.CCR-12-0372To request permission to re-use all or part of this article, use this link



http://clincancerres.aacrjournals.org/lookup/doi/10.1158/1078-0432.CCR-12-0372



http://clincancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://clincancerres.aacrjournals.org/content/early/2012/09/16/1078-0432.CCR-12-0372


Documents

Characterization of a Large Panel of Patient-Derived Tumor … · 2012. 9. 16. · Cancer Therapy: Preclinical See commentary by Kopetz et al., p. 5161 Characterization of a Large