RESOURCE ARTICLE

A novel patient-derived orthotopic xenograft model of esophagealadenocarcinoma provides a platform for translational discoveriesOmkara Lakshmi Veeranki1, Zhimin Tong1, Alicia Mejia1, Anuj Verma2, Riham Katkhuda2, Roland Bassett3,Tae-Beom Kim4, Jing Wang4, Wenhua Lang2, Barbara Mino2, Luisa Solis2, Charles Kingsley5, William Norton6,Ramesh Tailor7, Ji Yuan Wu8, Sunil Krishnan9, Steven H. Lin9, Mariela Blum8, Wayne Hofstetter10,Jaffer Ajani8, Scott Kopetz8 and Dipen Maru1,2,*

ABSTRACTMouse models of gastroesophageal junction (GEJ) cancer strive torecapitulate the intratumoral heterogeneity and cellular crosstalkwithin patient tumors to improve clinical translation. GEJ cancersremain a therapeutic challenge due to the lack of a reliable mousemodel for preclinical drug testing. In this study, a novel patient-derivedorthotopic xenograft (PDOX) was established from GEJ cancer viatransabdominal surgical implantation. Patient tumor was compared tosubcutaneously implanted patient-derived tumor xenograft (PDX) andPDOX by Hematoxylin and Eosin staining, immunohistochemistryand next-generation sequencing. Treatment efficacy studies ofradiotherapy were performed. We observed that mechanicalabrasion of mouse GEJ prior to surgical implantation of a patient-derived tumor in situ promotes tumor engraftment (100%, n=6).Complete PDOX engraftment was observed with rapid intra- andextraluminal tumor growth, as evidenced by magnetic resonanceimaging. PDOXs contain fibroblasts, tumor-associated macrophages,immune and inflammatory cells, vascular and lymphatic vessels.Stromal hallmarks of aggressive GEJ cancers are recapitulated in aGEJ PDOX mouse model. PDOXs demonstrate tumor invasion intovasculature and perineural space. Next-generation sequencingrevealed loss of heterozygosity with very high allelic frequency inNOTCH3, TGFB1, EZH2 and KMT2C in the patient tumor, thesubcutaneous PDX and the PDOX. Immunohistochemical analysis ofHer2/neu (also known as ERBB2), p53 (also known as TP53) and p16(also known as CDKN2A) in PDX and PDOX revealed maintenance ofexpression of proteins found in patient tumors, but membranous EGFR

overexpression in patient tumor cells was absent in both xenografts.Targeted radiotherapy in this model suggested a decrease in size by61% according to Response Evaluation Criteria in Solid Tumors(RECIST), indicating a partial response to radiation therapy. Our GEJPDOX model exhibits remarkable fidelity to human disease andcaptures the precise tissue microenvironment present within the localGEJ architecture, providing a novel tool for translating findings fromstudies on human GEJ cancer. This model can be applied to studymetastatic progression and to develop novel therapeutic approachesfor the treatment of GEJ cancer.

This article has an associated First Person interview with the firstauthor of the paper.

KEY WORDS: Patient-derived orthotopic xenograft mouse model,Esophageal adenocarcinoma, Gastroesophageal junction cancer,Mouse models, Microenvironment

INTRODUCTIONSince the 1970s, gastroesophageal (GE) cancers are an increasingcause of cancer burden globally (Blot et al., 1991; Fitzmaurice et al.,2015). Despite recent treatment advances, including the use ofneoadjuvant chemoradiation therapy (Mokdad et al., 2018),localized esophageal and/or gastroesophageal junction (GEJ)adenocarcinoma remains a highly aggressive cancer with 5-yearsurvival rates of less than 30% (Fitzmaurice et al., 2015). One factorlimiting improvements in treatment efficacy is a lack of preclinicalmodels that recapitulate the growth pattern and radiological andpathological characteristics of patient tumors and that can be used toassess response to chemoradiation. Conventional preclinical modelsdo not entirely recapitulate the human disease and hamperdevelopment of new treatments. Cell line xenografts implantedsubcutaneously or orthotopically lack the architectural complexityof patient tumors, which include intratumoral inflammatory cells,blood vessels and the presence of stromal elements (Becher andHolland, 2006). Thus, genetically engineered mouse models(GEMMs) (Becher and Holland, 2006) and patient-derivedxenografts (PDXs) have emerged as promising translationalplatforms due to their resemblance to human tumors (Hidalgoet al., 2014; Siolas and Hannon, 2013). Well-established GEMMsof GEJ cancers (HK-ATPase hIL-1β, IL-1β;NFκBEGFP, IL-1β;Rag2−/−, Villin-cre,Smad4F/F,Trp53F/F, Cdh1F/wt and Pdx1-cre,Trp53F/F, Cdh1F/F) (Park et al., 2015; Tu et al., 2008) havecontributed significantly to our understanding of GEJ tumorbiology. Owing to the histological differences between mouse andhuman GEJ, GEMMs for GE cancers have been extremely difficultto develop (Tétreault, 2015). In the human GEJ, the nonkeratinizingesophageal squamous epithelium is formed between the distalReceived 20 June 2019; Accepted 8 November 2019

1Department of Pathology, The University of Texas MD Anderson Cancer Center,Houston, TX 77030, USA. 2Department of Translational Molecular Pathology,The University of Texas MD Anderson Cancer Center, Houston, TX 77030, USA.3Department of Biostatistics, The University of Texas MD Anderson Cancer Center,Houston, TX 77030, USA. 4Department of Bioinformatics and ComputationalBiology, The University of Texas MD Anderson Cancer Center, Houston, TX 77030,USA. 5Department of Imaging Physics, The University of Texas MD AndersonCancer Center, Houston, TX 77030, USA. 6Department of Veterinary Medicine andSurgery, The University of Texas MD Anderson Cancer Center, Houston, TX 77030,USA. 7Department of Radiation Physics, The University of Texas MD AndersonCancer Center, Houston, TX 77030, USA. 8Department of Gastrointestinal MedicalOncology, The University of Texas MD Anderson Cancer Center, Houston, TX77030, USA. 9Department of Radiation Oncology, The University of Texas MDAnderson Cancer Center, Houston, TX 77030, USA. 10Department of Thoracic andCardiovascular Surgery, The University of Texas MD Anderson Cancer Center,Houston, TX 77030, USA.

*Author for correspondence (dmaru@mdanderson.org)

D.M., 0000-0002-9134-7709

This is an Open Access article distributed under the terms of the Creative Commons AttributionLicense (https://creativecommons.org/licenses/by/4.0), which permits unrestricted use,distribution and reproduction in any medium provided that the original work is properly attributed.

1

© 2019. Published by The Company of Biologists Ltd | Disease Models & Mechanisms (2019) 12, dmm041004. doi:10.1242/dmm.041004

Disea

seModels&Mechan

isms

esophagus, composed of squamous mucosa, abruptly transitioningto gastric columnar mucosa. In mouse, this squamocolumnarjunction is keratinized and is within the stomach (Treuting et al.,2011). The keratinized epithelium in mouse GEJ epithelium rendersthe mice resistant to injuries and to the development of GE cancers.Moreover, transgenic mice such as p63−/− have limitations as arobust GEJ cancer model as they do not survive to adulthood (Wanget al., 2011). Although immunocompetent murine cancer modelsand GEMMs develop spontaneous de novo tumors with a co-evolving microenvironment and an intact immune system, thesesystems are biased towards mouse-specific efficacy.PDX models have become invaluable in translational cancer

research because they retain a patient’s tumor characteristics andtumor heterogeneity, and predict sensitivity to treatment better thancell-line xenograft models. They preserve tumor heterogeneity ofparental tumors at histological and molecular levels even aftermultiple passages in mice (Choi et al., 2016; Dodbiba et al., 2015).PDXs are valuable tools for personalized medicine despite humanstroma being replaced by murine stroma. PDXs at subcutaneous andother heterotopic sites are attractive because of their technicalsimplicity. However, the subcutaneous sites differ from the tubulargastrointestinal tract in terms of anatomical boundaries and the tumormicroenvironment, limiting the applicability of subcutaneous PDXmodels for understanding tumor growth and response to treatment.Further, aggressive GEJ cancers are characterized by rich vasculature,lymphatic vessels, increased tumor-associated macrophages (TAMs)and cancer-associated fibroblasts. Subcutaneous PDX limits theusefulness of the PDX model in metastatic, angiogenic andtumor microenvironment studies. To overcome this limitation, aGEJ patient-derived orthotopic xenograft (PDOX)was established bydirectly implanting the GE cancer at the mouse GEJ, which enabledtumor growth similar to that of human tumors.

RESULTSEstablishing a PDOX mouse model of GEJ cancerPDOX models are rapidly gaining popularity, as they have a longhistory of mimicking patient tumor growth andmetastasis in cancersof the colon, pancreas, ovary, lung and stomach (Furukawa et al.,1993b; Hoffman, 2015). Unlike other sites in the tubulargastrointestinal tract (Furukawa et al., 1993a,b), no PDOX modelsare available for esophageal and/or GEJ cancers. Therefore, weestablished an esophageal/GEJ PDOX model by transabdominallyimplanting a biopsy sample of esophageal adenocarcinoma at thedistal esophagus/GEJ of six female SCID mice (Fig. S1A-I).Magnetic resonance imaging (MRI) with and without a gadolinium-based contrast agent (Magnevist) was used to visualize tumorboundaries and characterize tumor growth. T1-weighted imageswith contrast clearly showed enhancement of the tumor boundaries(Fig. S1J,K) and regions of hyperintensity and hypointensity,indicating a mixture of viable and necrotic tumor tissue. MRIstudies conducted 7 days after surgical implantation demonstrated100% (n=6) engraftment (Fig. S1I). MRI also showed an increasein the primary tumor size at each evaluated time point, indicatingthat the GEJ PDOXs grew over time (Fig. 1A). The mean tumorvolume (±s.e.m.) was 5.2±1.1 mm3 at Day 14 after implantation,4.3±1.2 mm3 at Day 21, 21.8±10.9 mm3 at Day 28, 18.6±12.3 mm3

at Day 49 and 84.5±4.6 mm3 at Day 87.

The PDOX mouse model mimics the patient GEJ tumor athistological and molecular levelThe histopathological features of the PDOX tumors were verysimilar to those of the patient tumor used to generate the PDOX

(Fig. 1B-F). The PDOXs exhibited small glandular and single-cellarchitecture with intracellular mucin (Fig. 1D), as did the patienttumor (Fig. 1B). An invasive growth pattern with transmuralextension to the mucosa and mucosal ulceration (Fig. 1D) resembledthe pattern observed on esophagogastroduodenoscopy in the patient.The tumor stroma was characterized by desmoplastic changesand mononuclear inflammatory infiltrates. Immunohistochemicalanalysis using an antibody specific to the CD68+ human epitopedepicted TAMs in patient tumor alone and not in subcutaneous PDXand PDOX (Fig. S2). Clinically important tumor microenvironmenthallmarks of aggressive behavior such as perineural (Fig. 1E) andintravascular invasion (Fig. 1F) were also observed. Lymphovascularinfiltration in the GEJ PDOX was depicted by CD31 (also known asPECAM1) immunohistochemistry (Fig. 1F).

Loss of heterozygosity (LOH)with very high allelic frequencywasobserved in NOTCH3, TGFB1, EZH2 and KMT2C in the patienttumor, the subcutaneous PDX and the PDOX. An additional LOHevent was observed in major histocompatibility complex, class IA(HLA-A) in the subcutaneous PDX and PDOX, but not in the patienttumor (Fig. 1G; Table S1), raising the possibility of the evolution of atumor cell clone(s) with expression of neoantigens in xenografts. ThePDOX and subcutaneous PDX exhibited somatic single-nucleotidevariants (SNVs) inARID1A (in-frame loss of codon, allelic frequency34%and47%) thatwere not observed in the patient tumor.Additionallow-frequency SNVs in KDM6A (in-frame gain of codon, allelicfrequency 7%) and XPO1 (splice region variant, allelic frequency6%) were identified in the PDOX, but not in the patient tumor or thesubcutaneous PDX (Fig. 1H; Table S2). The frequency of SNVs andLOH found in the PDOX but not in the patient tumor was consistentwith expectations for subsequent generations of a tumor.

Tumor cells in parental tumor, subcutaneous PDX and PDOX hada similar pattern of staining for Her2/neu (also known as ERBB2),p53 (also known as TP53) and p16 (also known as CDKN2A) byimmunohistochemistry (Fig. 2 and Table 1). Her2/neu expressionwas absent (score 0 of 3) in parental tumor, subcutaneous PDX andPDOX (Fig. 2A). Nuclear p53 staining was observed in 60% or moretumor cells in the parental tumor andwas retained in the subcutaneousPDX and PDOX (Fig. 2B). Nuclear or cytoplasmic p16 in the tumorcells was not seen in the patient tumor or in subcutaneous PDX andPDOX (Fig. 2C). Interestingly, membranous staining for EGFR wasobserved in more than 95% of the tumor cells of the patient tumor,whereas staining was absent in all the cells of subcutaneous PDX andPDOX (Fig. 2D).

The PDOX mouse model mimics the GEJ tumor growthpattern in patient tumorTumor growth kinetic studies revealed continuous tumor spread(Figs 1A and 3A) to the intraluminal and extraluminal esophagealwall and stomach wall and discontinuous spread in theperiesophageal and perigastric soft tissue, mimicking the localgrowth patterns of esophageal adenocarcinoma in patients.Discontinuous multifocal lesions were observed within the liningof the gastric wall at Day 200 after implantation (i.e. Day 82 afterirradiation), suggesting persistent growth of the tumor (Fig. 3B,C).

The PDOX mouse model recapitulates response to radiationin GEJ cancer patientsTo determine how well the PDOX model recapitulated the effects ofradiation treatment in patients, we administered a radiation dose of10 Gy to the PDOX tumors. The radiation treatment field (Fig. 3D;Fig. S3) was comparable to that in a typical treatment plan delivered topatients. The radiation dose was targeted at the tumor in the distal

2

RESOURCE ARTICLE Disease Models & Mechanisms (2019) 12, dmm041004. doi:10.1242/dmm.041004

Disea

seModels&Mechan

isms

esophagus and GEJ using a 5-mm-thick Wood’s metal (Cerrobend)block with a 2.0 cm×1.5 cm window placed directly above the GEJ(Fig. S3). Both the Response Evaluation Criteria in Solid Tumors(RECIST) and tumor volume measurements showed a decrease in thetumor burden in the PDOX model 28 days after radiation treatment(Fig. 3E). Radiation-treated tumors decreased in size by 61%according to RECIST, indicating a partial response to radiation

therapy. At necropsy, analysis of tumor volumemeasurements showeda decrease of 49% (68.3±26.9 mm3) at 30 days after irradiation. Thegross pathological volume of the esophageal/GEJ PDOX tumors waslinearly correlated with both the volumetric measurements (r=0.9,P<0.001) and RECIST (r=0.9, P<0.001). On autopsy, no significantradiotoxicity was noted in the GEJ, the surrounding soft tissues, lungand liver by macroscopic and histopathological examination.

Fig. 1. The PDOXmousemodel of esophageal adenocarcinomamimics the tumor growth pattern, and histopathological andmolecular characteristicsof the patient tumor. (A) MRI (coronal view) demonstrating growth of esophageal/GEJ PDOX (arrows) in two mice at day (D) 14, D21, D28, D49 and D87. Inmouse #1, at D49, there is clear evidence of intra- and extraluminal spread in the esophageal wall. (B-F) Histopathological evaluation by Hematoxylin and Eosin(H&E) staining and immunohistochemistry at 200× magnification: (B) glandular architecture (arrow) and cytologic features of the patient’s tumor; (C) glandulararchitecture and cytologic features of subcutaneously implanted PDX; (D) PDOX exhibiting glands and cellular features (thin arrow) similar to the histologicfeatures of the patient tumor (squamous epithelium of esophagus, thick arrow); (E) a focus of perineural invasion in PDOX (nerve bundle, thin arrow; perineuraltumor, thick arrow); (F) immunohistochemistry of CD31 depicting lymphovascular infiltration (arrow) in GEJ PDOX. (G,H) Next-generation sequencing resultsdemonstrating allelic frequency (%) for genes with loss of heterozygosity (LOH) (G) and somatic mutation (H).

3

RESOURCE ARTICLE Disease Models & Mechanisms (2019) 12, dmm041004. doi:10.1242/dmm.041004

Disea

seModels&Mechan

isms

DISCUSSIONIn summary, we generated a novel orthotopic mouse model ofesophageal/GEJ adenocarcinoma and demonstrated effectivedelivery of radiation to this PDOX model. The orthotopic mousemodel of esophageal/GEJ adenocarcinoma mimics thebioavailability of radiation and likely chemotherapy or targetedtherapy in patients (Malaney et al., 2014; Stewart et al., 2015; Zhanget al., 2013). The PDOX models are more suitable for co-clinicaltrials as they represent a next-generation approach to optimizingoncology drug development by a more profound and timelyunderstanding of tumor response to experimental agents, whilehelping to advance clinical development plans. Given thesimilarities in the imaging and pathologic characteristics of thisPDOX model and the patient’s tumor in the esophagus/GEJ, inaddition to its utility in metastatic, angiogenic and tumormicroenvironment studies, we conclude that this model is betterthan subcutaneous PDX models for studying tumor biology.Expression of CD68 (a pan-macrophage marker specific forhuman antigen) was found to be positive only in the patient tumorstroma and not in the subcutaneous PDX and PDOX, suggestingthat the TAMs in PDX and PDOX may have been replaced by

mouse macrophages. Loss of EGFR overexpression on tumor cellsobserved in subcutaneous PDX and PDOX indicated EGFRinactivation in the PDX model for this particular tumor. Previousreports suggest that EGFR amplification seen in patient tumors isoften lost in cell culture but not in xenografts developed from patienttumors (Pandita et al., 2004; Stockhausen et al., 2011). Since theprimary antibody used to stain EGFR was human specific, it isunlikely that absence of EGFR in tumor cells in xenografts is due totechnical issues or antibody clone. Therefore, the possibility of anew tumor cell clone with loss of EGFR expression in xenograftscannot be excluded. This finding needs to be tested in a largersample size to determine the presence and significance of EGFRloss in PDXs compared to the parental tumor. Although our modeldid not demonstrate distant metastasis, metastasis developing from aPDOXmodel like this onewould bemore representative of a stage-IVesophageal adenocarcinoma than would any other available models.Further, no significant radiotoxicities, such as esophagitis, gastritis,pneumonitis and inflammation, in liver adjacent to the targeted tumormass indicate that the GEJ PDOX model has applications inradiotherapy-based studies for concurrent examination of tumorresponses to radiation and radiation damage to normal tissues. The

Fig. 2. Immunohistochemical characterization of the PDOX mouse model of esophageal adenocarcinoma compared to patient tumor.(A) Immunohistochemistry staining for Her2/neu was negative in all tumor cells. (B) p53 demonstrated nuclear staining across all tumor cells. (C) Tumor cellswere negative for p16 staining with weak to moderate stromal staining in PDX and PDOX. (D) EGFR membranous staining present in patient tumor cellswas lost in PDX and PDOX. Magnification, ×200.

4

RESOURCE ARTICLE Disease Models & Mechanisms (2019) 12, dmm041004. doi:10.1242/dmm.041004

Disea

seModels&Mechan

isms

limitations of this study include few tumor samples from two patients,as only a portion of PDX can be established. Tumors from bothpatients were engrafted without any complications from the surgicalimplantation. Further optimization and expansion of this model frommultiple patients for understanding metastatic spread, circulatingtumor markers and the efficacy of novel therapeutic agents iswarranted. The orthotopic mouse model of esophageal/GEJadenocarcinoma makes it possible to test multiple therapeuticoptions for prioritizing patient treatment, which is not possible tocarry out in a clinical setting. Despite inherent challenges, PDOXsystems are likely to become increasingly useful tools as theyclosely recapitulate the heterogeneity of human tumors and couldfacilitate more-efficient oncology drug development.

MATERIALS AND METHODSFresh tissue samples from surgically excised cutaneous metastasis from apatient with distal esophageal/GEJ adenocarcinoma were utilized togenerate the xenografts. De-identified patient samples were obtained withinformed consent and xenografted as per Institutional Review Board (IRB)-and Institutional Animal Care and Use Committee (IACUC)-approvedprotocols (LAB-04-0979 and IACUC-00001501-RN01).

Generation of subcutaneous PDXPatient tumor was minced into ∼3-mm3 pieces within 10 min of removalfrom the patient. The minced tumor pieces were immersed in Matrigel(#356230, Corning, Bedford, MA, USA). Ectopic PDXs of GEJ cancer weregenerated by surgically implanting the tumor pieces into the flanks of sixfemale (4 weeks old) SCID or athymic mice, as previously described(Rubio-Viqueira et al., 2006).

Establishment of PDOXSix-week-old female SCID mice were purchased from the Mouse Facility atour institute. For tumor implantation, the mice were anesthetized using 2%isoflurane and placed in the dorsal recumbent position with a heating padbeneath them. Themicewere administered buprenorphine (0.05-0.10 mg/kgintraperitoneally immediately before surgery and postoperatively every6-12 h for up to 72 h), and the fur over the ventral abdomen was clipped anddisinfected (Fig. S1A). A ventral midline incision of ∼1.5 cm was madefrom the mid-abdominal region to the xiphoid process. The linea alba wasincised similarly (Fig. S1B). The margins of the incisions were retracted,and the liver lobes were gently reflected with a moist cotton swab to allowvisualization of the stomach and distal esophagus (Fig. S1C). The stomachwas then lifted extracorporeally by placing traction on the greater

curvature with forceps. A gavage needle was positioned underneath thedistal esophagus to elevate it (Fig. S1D). Mechanical abrasion was thenapplied to the distal esophagus/GEJ with rat-tooth forceps (Fig. S1E),and a 1-mm3 minced tumor dipped in Matrigel was sutured to theesophagus/GEJ with a 6-0 polyglactin 910 suture (Fig. S1F). Afterimplantation of the 1-mm3 PDX, the surgical site was observed for anybleeding. Once hemostasis was confirmed, the displaced organs weregently repositioned, and the linea was closed with a 4-0 absorbablesuture. The skin was subsequently closed with wound clips (Fig. S1G).Ringer’s lactate solution (1 ml) was administered subcutaneously to preventdehydration (Fig. S1H). The mice were followed up daily, weighed twotimes per week and monitored for toxicity.

MRIMice bearing PDOXs were anesthetized with 3% isoflurane forinduction and ∼2% isoflurane for maintenance. Animals were placedonto the bed of a 7-T BioSpec MRI scanner with a 30-cm bore (BrukerBioSpin, Billerica, MA, USA). Two-dimensional (2D) respiratory-gatedT2-weighted rapid acquisition with relaxation enhancement (RARE)images were acquired. Two-dimensional T1-weighted multi-slice multi-echo pre- and postcontrast images (Magnevist, Bayer, Whippany, NJ,USA) were acquired in the sagittal plane, and 2D T1-weightedfast-spoiled-gradient-echo pre- and postcontrast images were acquiredin the axial plane.

Radiation treatment to the PDOXThe PDOX-bearing mice were irradiated in a single 10-Gy fraction with a250-kVp x-ray beam (15 mA, 0.35 mm copper filter) from a Philips RT 250orthovoltage x-ray machine. An anesthetized mouse was placed inside aconforming cavity in a rubbery synthetic gel (Superflab) phantom placed ona 15-cm-thick acrylic block providing full backscatter. The field size was5×5 cm, the source-to-skin distance was set at 50.0 cm, and a 5-mm-thickWood’s metal (Cerrobend) block with a 2.0×1.5-cm window was placeddirectly over the GEJ. The dose output in this irradiation geometry was74.3 cGy/min (Fig. S2).

Radiation response evaluationRECIST version 1.1 was used to evaluate tumor response to therapy(Eisenhauer et al., 2009) Target lesions were chosen at baseline on thebasis of modified RECIST criteria. These same target lesions wereexamined for the 2D volumetric analysis, which was conducted usingImageJ (Dello et al., 2007).

Postmortem examination and histological analysesAfter the mice were euthanized, the full extent of tumor growth wasdetermined by autopsy. Gross examinations of the PDOX lesions and organssuch as the lungs and liver were performed to assess the extent oflocoregional spread. The PDOX tumors were measured using verniercalipers to determine their volume for comparison with the MRI volumetricmeasurements.

The tumors were measured in situ and excised, fixed with 4%formaldehyde, processed and embedded in paraffin blocks. Tumorsections (5 µm) were cut with a microtome (Microm HM355S; ThermoFisher Scientific, Waltham, MA, USA). H&E, CD31 (#77699; CST,Denvers, MA, USA), Her2/neu (#790-2991; Roche, Santa Clara, CA,USA), p53 (PA0057; Leica Microsystems, Buffalo Grove, IL, USA), p16(#705-4713; Roche, Santa Clara, CA, USA), EGFR (MAB13265; Abnova,Walnut, CA, USA) and CD68 (ab213363; Abcam, Cambridge, MA, USA)staining and validation were performed according to standard protocols in aClinical Laboratory Improvement Amendments (CLIA)-designatedlaboratory following College of American Pathologists guidelines (Delloet al., 2007). A detailed antibody list can be found in Table S3. Slides werescanned at magnifications of 100× and 200×. Images of the slides werecaptured with a light microscope (ColorView I, BX43F; Olympus, Tokyo,Japan). A pathologist reviewed the stained sections to determine thehistopathological features of the patient tumor and the xenograft tumors.

Table 1. Immunohistochemical staining intensity (0-3) and percentagepositive tumor cells in the PDOX mouse model of esophagealadenocarcinoma compared to patient tumor

Immunohistochemistrymarker and stainingintensity

Patienttumor (%)

SubcutaneousPDX (%)

PDOX(%)

Her2/neu 0 100 100 1001+ 0 0 02+ 0 0 03+ 0 0 0

p53 0 0 40 01+ 0 20 02+ 20 25 803+ 80 15 20

p16 0 100 100 1001+ 0 0 02+ 0 0 03+ 0 0 0

EGFR 0 5 99 1001+ 10 1 02+ 70 0 03+ 15 0 0

5

RESOURCE ARTICLE Disease Models & Mechanisms (2019) 12, dmm041004. doi:10.1242/dmm.041004

Disea

seModels&Mechan

isms

Next-generation sequencingA focus of tumor with higher than 70% tumor cellularity was marked on anH&E-stained section. Matching areas of the tumor were macrodissected forDNA sequencing. Normal squamous epithelium from the excised skin wasused as a matched control. Briefly, we used a KAPA HyperPrep librarypreparation kit (Kapa Biosystems, Wilmington, MA, USA) to prepareindexed libraries from 200 ng genomic DNA that had been sheared using aBiorupter Ultrasonicator (Diagenode, Denville, NJ, USA). The multiplexedlibrary pool was hybridized to the MD Anderson T200.1 probe pool (RocheNimbleGen, Madison, WI, USA). Following hybridization and reactioncleanup, the enriched libraries were amplified with seven cycles ofpostcapture PCR, then assessed for target enrichment by quantitativePCR. Sequencing was performed in one lane of a HiSeq4000 Sequencer(Illumina, San Diego, CA, USA) using a 150 bp paired-end configuration.

Bioinformatic analysisA customized hot spot somatic mutation bioinformatics pipeline focusing onanalysis of 200 (T200) most commonly mutated genes was used. Mousesequences were filtered out at the alignment. We aligned the T200 target-capture deep-sequencing data to human reference assembly hg19 usingBWA (Li and Durbin, 2009) and removed duplicate reads using Picard

(DePristo et al., 2011). We called SNVs and small indels using an in-house-developed analysis pipeline (Zhou et al., 2014) that classified variants intotwo categories – somatic and LOH – on the basis of variant allelefrequencies in the tumor and matched normal tissues.

AcknowledgementsWe thank Mayur Virarkar for enhancing radiograph images and providing MRItumor measurements, Kim-Anh Vu for helping with figures and Amy Ninetto forediting the manuscript.

Competing interestsThe authors declare no competing or financial interests.

Author contributionsConceptualization: O.L.V., Z.T., D.M.; Methodology: O.L.V., A.M., A.V., R.K., R.B.,T.-B.K., J.W., W.L., B.M., L.S., C.K., W.N., R.T., J.Y.W., S. Krishnan, S.H.L., M.B., W.H.,J.A., S. Kopetz, D.M.; Software: A.V., R.K., T.-B.K., J.W., W.L., B.M., C.K., D.M.;Validation: O.L.V., A.M., R.K., J.W., W.L., B.M., L.S., R.T., M.B., W.H., J.A., D.M.; Formalanalysis:O.L.V., Z.T.,A.V., T.-B.K., J.W., L.S.,C.K.,W.N.,D.M.; Investigation:O.L.V.,A.V.,T.-B.K., C.K., W.N., R.T., J.Y.W., S.H.L., M.B., W.H., J.A., S. Kopetz, D.M.; Resources:T.-B.K., C.K., W.N., S. Krishnan, S.H.L., W.H., J.A., S. Kopetz, D.M.; Data curation:O.L.V., Z.T., T.-B.K.; Writing - original draft: O.L.V.; Writing - review & editing: O.L.V.,

Fig. 3. Growth kinetics andresponse to targeted radiotherapyin the esophageal adenocarcinoma/GEJ PDOXmousemodel. (A) Growthrate of PDOX over time. The PDOXswere irradiated at D114 afterimplantation using a 250-kVp x-raybeam at a dose of 10 Gy. At D82 afterradiation, multifocal lesions were foundwithin the esophageal and gastric wall(red arrow). (B,C) MRI images showingdiscontinuous spread of GEJ PDOXwithin the esophageal and gastric wallon coronal (B) and axial (C) view. Thethick arrows indicate the primaryPDOX; the thin arrows indicatediscontinuous lesions. (D) Treatmentalgorithm for radiation studies. ThePDOXs were irradiated (10 Gy) onD0. (E) Evaluation of response toradiotherapy by MRI at D0, D7 andD28 (arrows indicate the PDOX).

6

RESOURCE ARTICLE Disease Models & Mechanisms (2019) 12, dmm041004. doi:10.1242/dmm.041004

Disea

seModels&Mechan

isms

D.M.; Supervision: O.L.V., A.M., S. Krishnan, S.H.L., S. Kopetz, D.M.; Projectadministration:O.L.V.,A.M.,A.V.,D.M.;Fundingacquisition:D.M.,O.L.V.,S.Kopetz,C.K.

FundingThis work was supported by an Independent Investigator Award from the CancerPrevention and Research Institute of Texas (RP140515 to D.M.), the U.S.Department of Defense (CA160928 to O.L.V.) and the National Institutes of Health/National Cancer Institute (CA184843 to S. Kopetz). The Small Animal ImagingFacility and Sequencing and Microarray Facility at The University of Texas MDAnderson Cancer Center are supported by the National Institutes of Health/NationalCancer Institute through Cancer Center Support Grant P30CA016672.

Supplementary informationSupplementary information available online athttp://dmm.biologists.org/lookup/doi/10.1242/dmm.041004.supplemental

ReferencesBecher, O. J. and Holland, E. C. (2006). Genetically engineered models haveadvantages over xenografts for preclinical studies. Cancer Res. 66, 3355-3358,discussion 3358-9. doi:10.1158/0008-5472.CAN-05-3827

Blot, W. J., Devesa, S. S., Kneller, R. W. and Fraumeni, J. F. Jr. (1991). Risingincidence of adenocarcinoma of the esophagus and gastric cardia. JAMA 265,1287-1289. doi:10.1001/jama.1991.03460100089030

Choi, Y. Y., Lee, J. E., Kim, H., Sim, M. H., Kim, K.-K., Lee, G., Kim, H.-I., An, J. Y.,Hyung, W. J., Kim, C.-B. et al. (2016). Establishment and characterisation ofpatient-derived xenografts as paraclinical models for gastric cancer. Sci. Rep. 6,22172. doi:10.1038/srep22172

Dello, S. A. W. G., van Dam, R. M., Slangen, J. J. G., van de Poll, M. C. G.,Bemelmans, M. H. A., Greve, J. W. W. M., Beets-Tan, R. G. H., Wigmore, S. J.and Dejong, C. H. C. (2007). Liver volumetry plug and play: do it yourself withImageJ. World J. Surg. 31, 2215-2221. doi:10.1007/s00268-007-9197-x

DePristo, M. A., Banks, E., Poplin, R., Garimella, K. V., Maguire, J. R., Hartl, C.,Philippakis, A. A., del Angel, G., Rivas, M. A., Hanna, M. et al. (2011). Aframework for variation discovery and genotyping using next-generation DNAsequencing data. Nat. Genet. 43, 491. doi:10.1038/ng.806

Dodbiba, L., Teichman, J., Fleet, A., Thai, H., Starmans, M. H. W., Navab, R.,Chen, Z., Girgis, H., Eng, L., Espin-Garcia, O. et al. (2015). Appropriateness ofusing patient-derived xenograft models for pharmacologic evaluation of noveltherapies for esophageal/gastro-esophageal junction cancers. PLoS ONE 10,e0121872. doi:10.1371/journal.pone.0121872

Eisenhauer, E. A., Therasse, P., Bogaerts, J., Schwartz, L. H., Sargent, D., Ford,R., Dancey, J., Arbuck, S., Gwyther, S., Mooney, M. et al. (2009). Newresponse evaluation criteria in solid tumours: revised RECIST guideline (version1.1). Eur. J. Cancer 45, 228-247. doi:10.1016/j.ejca.2008.10.026

Fitzmaurice, C., Dicker, D., Pain, A., Hamavid, H., Moradi-Lakeh, M., MacIntyre,M. F., Allen, C., Hansen, G., Woodbrook, R., Wolfe, C. et al. (2015). The globalburden of cancer 2013. JAMAOncol. 1, 505-527. doi:10.1001/jamaoncol.2015.0735

Furukawa, T., Fu, X., Kubota, T., Watanabe, M., Kitajima, M. and Hoffman, R. M.(1993a).Nudemousemetastaticmodelsofhumanstomachcancerconstructedusingorthotopic implantation of histologically intact tissue. Cancer Res. 53, 1204-1208.

Furukawa, T., Kubota, T., Watanabe, M., Kitajima, M. and Hoffman, R. M.(1993b). Orthotopic transplantation of histologically intact clinical specimens ofstomach cancer to nude mice: correlation of metastatic sites in mouse andindividual patient donors. Int. J. Cancer 53, 608-612. doi:10.1002/ijc.2910530414

Hidalgo, M., Amant, F., Biankin, A. V., Budinska, E., Byrne, A. T., Caldas, C.,Clarke, R. B., de Jong, S., Jonkers, J., Mælandsmo, G. M. et al. (2014).Patient-derived xenograft models: an emerging platform for translational cancerresearch. Cancer Discov. 4, 998-1013. doi:10.1158/2159-8290.CD-14-0001

Hoffman, R. M. (2015). Patient-derived orthotopic xenografts: better mimic ofmetastasis than subcutaneous xenografts. Nat. Rev. Cancer 15, 451-452. doi:10.1038/nrc3972

Li, H. and Durbin, R. (2009). Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25, 1754-1760. doi:10.1093/bioinformatics/btp324

Malaney, P., Nicosia, S. V. and Dave, V. (2014). Onemouse, one patient paradigm:new avatars of personalized cancer therapy. Cancer Lett. 344, 1-12. doi:10.1016/j.canlet.2013.10.010

Mokdad, A. A., Yopp, A. C., Polanco, P. M., Mansour, J. C., Reznik, S. I., Heitjan,D. F., Choti, M. A., Minter, R. R., Wang, S. C. and Porembka, M. R. (2018).Adjuvant chemotherapy vs postoperative observation following preoperativechemoradiotherapy and resection in gastroesophageal cancer: a propensity score-matched analysis. JAMAOncol. 4, 31-38. doi:10.1001/jamaoncol.2017.2805

Pandita, A., Aldape, K. D., Zadeh, G., Guha, A. and James, C. D. (2004).Contrasting in vivo and in vitro fates of glioblastoma cell subpopulations withamplified EGFR. Genes Chromosomes Cancer 39, 29-36. doi:10.1002/gcc.10300

Park, J. W., Park, D. M., Choi, B. K., Kwon, B. S., Seong, J. K., Green, J. E., Kim,D.-Y. and Kim, H. K. (2015). Establishment and characterization of metastaticgastric cancer cell lines frommurine gastric adenocarcinoma lacking Smad4, p53,and E-cadherin. Mol. Carcinog. 54, 1521-1527. doi:10.1002/mc.22226

Rubio-Viqueira, B., Jimeno, A., Cusatis, G., Zhang, X., Iacobuzio-Donahue, C.,Karikari, C., Shi, C., Danenberg, K., Danenberg, P. V., Kuramochi, H. et al.(2006). An in vivo platform for translational drug development in pancreaticcancer. Clin. Cancer Res. 12, 4652-4661. doi:10.1158/1078-0432.CCR-06-0113

Siolas, D. and Hannon, G. J. (2013). Patient-derived tumor xenografts:transforming clinical samples into mouse models. Cancer Res. 73, 5315-5319.doi:10.1158/0008-5472.CAN-13-1069

Stewart, E. L., Mascaux, C., Pham, N.-A., Sakashita, S., Sykes, J., Kim, L.,Yanagawa, N., Allo, G., Ishizawa, K., Wang, D. et al. (2015). Clinical utility ofpatient-derived xenografts to determine biomarkers of prognosis and mapresistance pathways in EGFR-mutant lung adenocarcinoma. J. Clin. Oncol. 33,2472-2480. doi:10.1200/JCO.2014.60.1492

Stockhausen, M.-T., Broholm, H., Villingshøj, M., Kirchhoff, M., Gerdes, T.,Kristoffersen, K., Kosteljanetz, M., Spang-Thomsen, M. and Poulsen, H. S.(2011). Maintenance of EGFR and EGFRvIII expressions in an in vivo and in vitromodel of human glioblastoma multiforme. Exp. Cell Res. 317, 1513-1526. doi:10.1016/j.yexcr.2011.04.001

Tetreault, M.-P. (2015). Esophageal cancer: insights from mouse models. CancerGrowth Metastasis 8, 37-46. doi:10.4137/CGM.S21218

Treuting, P. M., Dintzis, S. M., Liggitt, D. and Frevert, C. W. (2011). ComparativeAnatomy and Histology: A Mouse and Human Atlas. Academic Press.

Tu, S., Bhagat, G., Cui, G., Takaishi, S., Kurt-Jones, E. A., Rickman, B., Betz,K. S., Penz-Oesterreicher, M., Bjorkdhl, O., Fox, J. G. et al. (2008).Overexpression of interleukin-1beta induces gastric inflammation and cancerand mobilizes myeloid-derived suppressor cells in mice.Cancer Cell 14, 408-419.doi:10.1016/j.ccr.2008.10.011

Wang, X., Ouyang, H., Yamamoto, Y., Kumar, P. A., Wei, T. S., Dagher, R.,Vincent, M., Lu, X., Bellizzi, A. M., Ho, K. Y. et al. (2011). Residual embryoniccells as precursors of a Barrett’s-like metaplasia. Cell 145, 1023-1035. doi:10.1016/j.cell.2011.05.026

Zhang, X., Claerhout, S., Prat, A., Dobrolecki, L. E., Petrovic, I., Lai, Q., Landis,M. D., Wiechmann, L., Schiff, R., Giuliano, M. et al. (2013). A renewable tissueresource of phenotypically stable, biologically and ethnically diverse, patient-derived human breast cancer xenograft models. Cancer Res. 73, 4885-4897.doi:10.1158/0008-5472.CAN-12-4081

Zhou,W., Chen, T., Zhao, H., Eterovic, A. K., Meric-Bernstam, F., Mills, G. B. andChen, K. (2014). Bias from removing read duplication in ultra-deep sequencingexperiments. Bioinformatics 30, 1073-1080. doi:10.1093/bioinformatics/btt771

7

RESOURCE ARTICLE Disease Models & Mechanisms (2019) 12, dmm041004. doi:10.1242/dmm.041004

Disea

seModels&Mechan

isms