cxm: a new tool for mapping breast cancer risk in the ... · cxm: a new tool for mapping breast...

Microenvironment and Immunology

CXM: A New Tool for Mapping Breast Cancer Risk in theTumor Microenvironment

Michael J. Flister1,2, Bradley T. Endres1,2, Nathan Rudemiller2, Allison B. Sarkis1,2, Stephanie Santarriaga3,Ishan Roy4, Angela Lemke1,2, Aron M. Geurts1,2, Carol Moreno1, Sophia Ran5,6, Shirng-Wern Tsaih2,Jeffery De Pons1, Daniel F. Carlson7, Wenfang Tan8,9, Scott C. Fahrenkrug7,8,9, Zelmira Lazarova10,Jozef Lazar1,10, Paula E. North11, Peter S. LaViolette3, Michael B. Dwinell4, James D. Shull12,13,14, andHoward J. Jacob1,2,11

AbstractThemajority of causative variants in familial breast cancer remain unknown. Of the known risk variants, most are

tumor cell autonomous, and little attention has been paid yet to germline variants that may affect the tumormicroenvironment. In this study, we developed a system called the Consomic Xenograft Model (CXM) to mapgermline variants that affect only the tumormicroenvironment. In CXM, humanbreast cancer cells are orthotopicallyimplanted into immunodeficient consomic strains and tumormetrics arequantified (e.g., growth, vasculogenesis, andmetastasis). Because the strain backgrounds vary, whereas themalignant tumor cells do not, any observed changes intumor progression are due to genetic differences in the nonmalignant microenvironment. Using CXM, we definedgenetic variants on rat chromosome 3 that reduced relative tumor growth and hematogenous metastasis in the SS.BN3IL2Rg consomic model compared with the SSIL2Rg parental strain. Paradoxically, these effects occurred despite anincrease in the density of tumor-associated blood vessels. In contrast, lymphatic vasculature and lymphogenousmetastasis were unaffected by the SS.BN3IL2Rg background. Through comparative mapping and whole-genomesequence analysis, we narrowed candidate variants on rat chromosome 3 to six genes with a priority for futureanalysis. Collectively, our results establish the utility of CXM to localize genetic variants affecting the tumormicroenvironment that underlie differences in breast cancer risk. Cancer Res; 74(22); 6419–29. �2014 AACR.

IntroductionBreast cancer is the most prevalent female malignancy

(http://apps.nccd.cdc.gov/uscs/) and is highly heritable (1–4),

yet approximately 72% of breast cancer risk remains undefined(5). Heritable factors underlie most aspects of breast cancerrisk [e.g., incidence (5), age of onset (6), metastatic progression(7), and disease-free survival (8)]. In addition to variants thataffect tumor cells directly (i.e., tumorigenicity), heritabilityis implicated in multiple components of the tumor microen-vironment [e.g., tissue remodeling (9), angiogenesis (10, 11),and immunity (12)], which also affect tumorigenesis andprogression. However, the genetic variants underlying differ-ences in the tumor microenvironment have rarely been thefocus of genetic mapping studies and as such remain poorlydefined.

We have developed a new model of breast cancer (termedthe Consomic Xenograft Model; CXM) that focuses on geneticmapping of strain-specific variants that affect tumor progres-sion through the tumor microenvironment. A consomic rat isone in which an entire chromosome is introgressed into theisogenic background of another inbred strain by selectivebreeding (13). Thus, observed phenotypes can be linked tosingle chromosomes and then further elucidated by compar-ative sequence analysis and/or selective backcrossing to yieldsmaller congenics (13). In CXM, the consomic and parentalstrains are converted to SCID (severe combined immunode-ficiency), so that orthotopically xenografted human breastcancer cells can be tested in vivo. Because the human breastcancer cells are not varied between strains, any differencesin breast cancer progression (e.g., primary site growth,

1Human and Molecular Genetics Center, Medical College of Wisconsin,Milwaukee, Wisconsin. 2Department of Physiology, Medical College ofWisconsin, Milwaukee, Wisconsin. 3Department of Radiology, MedicalCollege ofWisconsin,Milwaukee,Wisconsin. 4Department ofMicrobiologyand Molecular Genetics, Medical College of Wisconsin, Milwaukee, Wis-consin. 5SimonsCooper Cancer Institute, Southern Illinois UniversitySchool of Medicine, Springfield, Illinois. 6Department of Medical Microbi-ology, Immunology, and Cell Biology, Southern Illinois University School ofMedicine, Springfield, Illinois. 7Recombinetics Inc, Saint Paul, Minnesota.8Department of Animal Science, University of Minnesota, Saint Paul,Minnesota. 9Center for Genome Engineering, University of Minnesota,Minneapolis, Minnesota. 10Department of Dermatology, Medical Collegeof Wisconsin, Milwaukee, Wisconsin. 11Department of Pediatrics, MedicalCollege of Wisconsin, Milwaukee, Wisconsin. 12McArdle Laboratory forCancer Research, University of Wisconsin, Madison, Wisconsin. 13Depart-ment of Oncology, School of Medicine and Public Health, University ofWisconsin, Madison, Wisconsin. 14UW Carbone Cancer Center, School ofMedicine and Public Health, University of Wisconsin, Madison, Wisconsin.

Note: Supplementary data for this article are available at Cancer ResearchOnline (http://cancerres.aacrjournals.org/).

Corresponding Authors: Howard J. Jacob, Human and Molecular GeneticsCenter, 8701WatertownPlankRd.,Medical College ofWisconsin,Milwaukee,WI 53226. Phone: 414-456-4887; Fax: 414-456-6516; E-mail: [email protected]; and Michael J. Flister, Human and Molecular Genetics Center, 8701Watertown Plank Rd., Medical College of Wisconsin, Milwaukee, WI 53226.Phone: 414-456-7534; Fax: 414-456-6516; E-mail: [email protected]

doi: 10.1158/0008-5472.CAN-13-3212

�2014 American Association for Cancer Research.

CancerResearch

www.aacrjournals.org 6419

on June 20, 2020. © 2014 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

Published OnlineFirst August 29, 2014; DOI: 10.1158/0008-5472.CAN-13-3212

http://cancerres.aacrjournals.org/

vasculogenesis, and distal metastasis) are due solely to geneticdifferences in the tumormicroenvironment, not themalignantcancer cells. CXM utilizes transgenically tagged human cancercells with defined properties (e.g., triple-negative, prometa-static, etc.). Thus, it enables testing of clinically relevant cancermodels in strain backgrounds with varying genetic predisposi-tions to breast cancer. Finally, we have developed CXM in therat, but postulate that this technique could also be applied tothe mouse and other model species.

For purposes of illustration, the basic concept of CXM isoutlined in Fig. 1. In brief, CXMutilizes transcription activator-like effector nucleases (TALEN; ref. 14) to generate SCID onany strain background by mutating the IL2Rg gene (15). Wefirst generated and characterized an IL2Rg-mutant parental SSstrain (SSIL2Rg), which was then selectively bred onto the SS.

BN3 consomic background to generate the SS.BN3IL2Rg con-somic. The parental SS (SS/JrHsdMcwi) rat strain is susceptibleto mammary tumors, whereas the BN (BN/NHsdMcwi) ratstrain is highly resistant to mammary tumors (16). Ourrationale for choosing SS (parental) and SS.BN3 consomic(i.e., BN chromosome 3 introgressed onto the isogenic SSbackground) was based on prior data demonstrating a 64%reduction in breast cancer incidence in the SS.BN3 consomiccompared with the SS (16) and other evidence that multipleoverlapping protective loci exist on rat chromosome 3 (16–19). We hypothesized that the protective effects of BNchromosome 3 could be in part due to changes in the tumormicroenvironment. To test this hypothesis, luciferase-taggedhuman MDA-MB-231 breast cancer cells (231Lucþ) wereorthotopically implanted and tracked for tumor growth,

Figure 1. Overview of the CXM. A,reduced susceptibility tomammarytumors (DMBA-induced) inSS.BN3consomic rats (n ¼ 14) comparedwith SS (n ¼ 40), as reportedpreviously by Adamovic andcolleagues (16). ��, P < 0.001, asdetermined by the Fisher exact testusing Bootstrap. B, schematicrepresentation of the SS and SS.BN3 genomes that were modifiedby TALEN-mediated editing of theIL2Rg gene. The numbered barsrepresent chromosomes that arederived from SS (white) or BN(black). Note that the only geneticdifferences between SSIL2Rg andSS.BN3IL2Rg are the inheritance ofchromosome 3 from the SS or BNrats. C, the transgenically labeledhuman 231Lucþ breast cancer cellswere orthotopically implanted intheMFP (indicatedbydashed lines)of SSIL2Rg and SS.BN3IL2Rg rats.Tumor progression (e.g., growth,vasculogenesis, and metastasis)were tracked over the duration ofthe experiment. D, because the231Lucþ tumor cells are the samebetween strains (gray), anydifferences in tumor progressioncan be attributed to differences inthe SSIL2Rg (green) and SS.BN3IL2Rg

(blue) microenvironments.

Flister et al.

Cancer Res; 74(22) November 15, 2014 Cancer Research6420




vasculogenesis, and distal metastasis. Using CXM, we dem-onstrate that genetic variants on rat chromosome 3 affectthe tumor microenvironment by causing blood vessel–spe-cific defects that attenuate tumor growth and decreasehematogenous metastasis, whereas lymphatic vasculatureand lymphogenous metastasis were unaffected. Finally, weused comparative analysis and whole-genome sequencing(WGS) of five rat strains (BN/JrHsdMcwi, Cop/Crl, F344/N,ACI/Eur, and SS/JrHsdMcwi) to prioritize cosegregatingcandidate genes on rat chromosome 3 for future analysisby gene editing or congenic mapping.

Materials and MethodsGeneration of SS.BN3IL2Rg and SSIL2Rg rat strainsAll procedures and protocols were approved by the Medical

College of Wisconsin (MCW, Milwaukee, WI) IACUC commit-tee. The IL2Rg gene was targeted in the SS/JrHsdMcwi rat byTALEN injection into single-cell rat embryos, as describedpreviously (20). Once established, a homozygous female ratfrom the SSIL2Rg line was intercrossed with a homozygous SS.BN3 male to yield heterozygous SS.BN3IL2Rg offspring (F1),followed by brother-sister mating to yield homozygous SS.BN3IL2Rg offspring by the F3 generation.

Tumor implantationLuciferase-tagged MDA-MB-231 (231Lucþ) cells were ortho-

topically implanted andmeasured as described previously (21),with slight modifications. Briefly, 231Lucþ cells (6� 106) in 50%Matrigel were orthotopically implanted into the mammary fatpads (MFP) of 4- to 6-week-old female SSIL2Rg (n ¼ 11) and SS.BN3IL2Rg (n¼ 19) rats. Tumor volumes were measured weeklyby calipers, as described previously (21).

Detection of lung metastasis by ex vivo luminescentimagingRats were intraperitoneally injected with 200 mL of

D-luciferin (Promega) diluted in sterile saline (40 mg/mL). Thesubstrate was allowed to circulate for 5 minutes beforerats were euthanized and lungs were removed for biolumines-cence imaging using the Xenogen Ivis Lumina (Caliper LifeSciences).

Detection of metastatic burden by luciferase activityIpsilateral axillary lymph nodes and lungs were excised,

washed in PBS, and homogenized in 0.2 and 2 mL of cold cellculture lysis reagent buffer (Promega), respectively. Celldebris was removed by centrifugation. Protein concentra-tions of cleared lysates were determined by Bradford assay(Bio-Rad). Fifty microliters of Luciferase Assay Reagent(Promega) was mixed with 10 mL of lysate, and a 10-secondaverage of luminescence was detected using a VarioskanFlash luminometer (ThermoFisher). Cell culture lysisreagent buffer without tissue homogenates and lymph nodesor lungs of non–tumor-bearing rats was used to calculatebackground and subtracted from the results. The net resultsare expressed as relative light units normalized per milli-gram of total protein.

Histologic detection of lung metastasisMetastatic lesions in the lungs of tumor-bearing SSIL2Rg

(n ¼ 5) and SS.BN3IL2Rg (n ¼ 5) rats were also assessed byhematoxylin and eosin (H&E) staining, as described previ-ously (22). Formalin-fixed lung sections were H&E stainedusing a Tissue-Tek Prisma Automated Stainer (Sakura)according to the manufacturer's protocol. Following H&Estaining, three serial sections per lung were examined by twoblinded observers.

Matrigel plug angiogenesis assayFemale SS (n ¼ 9) and SS.BN3 (n ¼ 19) rats at 6–8 weeks of

age were implanted in the MFP with 500 mL of 100% Matrigel(cat. CB-40234; BDBiosciences) supplementedwith 500 ng/mLof recombinant rat VEGFA164 purchased (R&D Systems). At 5days postimplantation, Matrigel plugs were excised, snap-frozen in optimum cutting temperature medium (Tissue Tek),and stained with anti-CD31 as described in the SupplementaryMaterials and Methods.

Blood and lymphatic vessel densities231Lucþ tumors and Matrigel plugs were excised, frozen-

sectioned, and immunostained with antibodies against theblood vessel marker, CD31, or the lymphatic vessel marker,LYVE-1, as described previously (23). To quantify tumor bloodvessel density (BVD) and lymphatic vessel density (LVD), 3 to 4representative images of CD31þ structures (BVD) and LYVE-1þ structures (LVD) per tissue were acquired at �100 magni-fication using a Nikon E400 microscope equipped with a SpotInsight camera (Nikon Instruments). BVD and LVD, averagenumber of vessels per area of the field � SEM (n ¼ 5–7 rats/group).

Quantification of BVD in DMBA-induced mammarytumors

Acquisition of 7,12-dimethylbenz(a)anthracene (DMBA)-induced mammary tumors from the SS.BN3 consomic (n ¼9) and SS (n ¼ 9) rats were previously described by Adamovicand colleagues (16). These tumors were previously formalin-fixed and paraffin-embedded, precluding us from using thetraditional anti-CD31 antibody that does not recognize for-malin-fixed antigens. To circumvent this issue, DMBA-inducedmammary tumors were stained with anti-SH2B3 antibodies(Sigma Aldrich; cat. HPA005483), which we have demonstratedto specifically recognize CD31þ blood vessels, but not LYVE1þ

lymphatic vessels (Supplementary Fig. S2). The area of SH2B3þ

blood vessels across the entire tumor cross section wasacquired at �100 magnification on a Nikon Eclipse 80i micro-scope (Nikon Instruments) equipped with an automated stageand aQImagingMicropublisher camera. Images were analyzedusing customMatlab code (MathWorks) that stitched togethertissue photographs (�100–750 per sample) and segmented thetissue into to differentiate positively stained vessels (Fig. 4C).The histology segmentation process included a white back-ground correction and contrast optimization. Static thresholdswere applied to each photo for vessel segmentation. Relativevascular density was then calculated for each sample andnormalized to the entire cross-sectional tumor area.

Genetic Determinants of Breast Cancer Risk

www.aacrjournals.org Cancer Res; 74(22) November 15, 2014 6421




Peripheral blood analysisWhole blood from anesthetized rats was analyzed by auto-

mated complete blood counts (Marshfield Laboratories) andflow cytometrywas performed, as described previously (24, 25).

Genomic sequencing and analysisGenomic sequence was accessed from the Rat Genome

Database (http://rgd.mcw.edu/) and has been described indetail elsewhere (26–28).

Statistical analysisStatistical analyses were performed using Sigma Plot 11.0

software. Data are presented as mean � SEM. All data wereanalyzed by an unpaired Student t test.

Detailed Materials and Methods are provided in the Sup-plementary Data.

ResultsConsomic xenograft model

Traditional mapping studies localize cancer risk variants(29), but do not differentiate between variants that affectmalignant cells directly versus those that affect the tumormicroenvironment. Here, our goal was to develop a newgeneticmodel of breast cancer risk (i.e., CXM), whereby geneticvariants that specifically affect breast cancer progressionthrough the tumor microenvironment are mapped.

Following strain generation (SSIL2Rg and SS.BN3IL2Rg), wefirst examined gross organmorphologies and peripheral bloodphenotypes to confirm SCID status and noted strain-depen-dent differences. No differences in viability, body weight, organweight and morphology, and some peripheral blood pheno-types (T-cells, B-cells, and NK cells) were observed betweenSSIL2Rg (n ¼ 7) and SS.BN3IL2Rg (n ¼ 11) rats (SupplementaryFig. S1). Differences were observed in circulating monocytes ofSS.BN3IL2Rg (0.4� 0.01 per mL blood, P < 0.001) compared withSSIL2Rg rats (0.17 � 0.02 per mL blood); in neutrophils of SS.BN3IL2Rg (1.9 � 0.3 per mL blood, P < 0.001) compared withSSIL2Rg rats (4.3 � 0.4 per mL blood); and in total WBCs of SS.BN3IL2Rg (2.4 � 0.4 per mL blood, P < 0.001) compared withSSIL2Rg rats (5.6 � 0.5 per mL blood; Supplementary Table S1).Compared with the immunocompetent parental SS and SS.BN3 strains, circulating lymphocytes were reduced 4- to 9-fold(P < 0.001) in SSIL2Rg and SS.BN3IL2Rg (SupplementaryTable S1), whereas neutrophils were increased by 5- to 9-fold(P < 0.001) in SS.BN3IL2Rg and SSIL2Rg compared with theparental strains (Supplementary Table S1). Similar changes inleukocytes have been reported in other immunodeficientmod-els [e.g., SCID rat (15) and SCID mouse (30)]. Taken together,these data demonstrate that SSIL2Rg and SS.BN3IL2Rg are bothSCID, with limited strain-dependent differences in immunity.

Tumor growthPreviously, we showed that the SS.BN3 consomic rat has

significantly reduced tumorigenicity in a DMBA-induciblemodel of breast cancer (16). We postulated that this could beat least partially due to genetic differences in the tumormicroenvironment, rather than inherent differences in the

malignant tumor cells only. To test this possibility, 231Lucþ

(6 � 106 cells) were orthotopically implanted in the MFP ofSSIL2Rg and SS.BN3IL2Rg rats and tumor growth was measuredby caliper measurement at 10, 17, and 24 days postimplanta-tion. Compared with 231Lucþ tumors in SSIL2Rg rats at 24 dayspost-implant (5,117� 1,038 mm3, n¼ 11), tumors grown in SS.BN3IL2Rg rats (2,616 � 624 mm3, n ¼ 19) were roughly 2-foldsmaller (P < 0.05; Fig. 2), suggesting that BN-derived antitumorvariants or SS-derived protumor variants reside on rat chro-mosome 3 and function through the tumormicroenvironment.

Tumor angiogenesis and hematogenous metastasisTo assess BVD, 231Lucþ tumors that were orthotopically

implanted inMFPs of SSIL2Rg (n¼ 5 rats) and SS.BN3IL2Rg (n¼ 6rats) rats were excised at 24 days postimplantation and stainedwith antibodies against the blood vessel marker, CD31 (23, 31)(Fig. 3A). Compared with tumor BVD in SSIL2Rg rats (101 � 11vessels/field), the BVD in SS.BN3IL2Rg tumors was significantlyincreased by 27% (130 � 18 vessels/field; P < 0.05; Fig. 3B).Despite increased BVD in SS.BN3IL2Rg tumors, we observedfewer blood vessels with open lumens in SS.BN3IL2Rg tumors(3.2 � 1.2%; P < 0.05) compared with tumors implanted inSSIL2Rg rats (6.2 � 2.0%; Fig. 3C), suggesting that although SS.BN3IL2Rg had a higher BVD, a greater percentage of thesevessels were collapsed and nonfunctional. A similar phenotype(i.e., high density of malformed vessels) was observed byPandey andWendell (32) in estrogen-induced pituitary tumorsfrom an overlapping F344.BN-Edpm3BN congenic rat with a75.7-Mb congenic interval (chr3:36.6-112.3 Mb), suggestingthat the vascular defects might be due to BN-derived variantsin a common region of rat chromosome 3.

We also observed a small percentage of CD31þ tumor bloodvessels being invaded by 231Lucþ tumor cells in both SSIL2Rg

and to a lesser degree in SS.BN3IL2Rg rats, demonstrating that

Figure 2. Growth curve of 231Lucþ breast carcinomacells implanted in SS.BN3IL2Rg and SSIL2Rg rats. The growth of 231Lucþ tumors was monitoredby caliper measurement at 10, 17, and 24 days postimplantation in SS.BN3IL2Rg (n ¼ 19) and SSIL2Rg (n ¼ 11) rats. Data are presented as meantumor volume� SEM. �,P < 0.05, as determined by the unpaired Studentt test.

Flister et al.





tumor cells were actively undergoing hematogenous metasta-sis. Quantification of CD31þ blood vessels invaded by 231Lucþ

cells revealed a significant decrease of vascular invasion in SS.BN3IL2Rg tumors (0.8� 0.5%; P < 0.001) compared with SSIL2Rg

(2.5 � 0.5%; Fig. 3D), suggesting that 231Lucþ have reducedmetastatic potential in the SS.BN3IL2Rg rat that is blood vesseldependent. Likewise, upon comparison of metastatic burdenin the lungs of tumor-bearing SSIL2Rg and SS.BN3IL2Rg rats, weobserved decreased metastatic lung nodules (Fig. 3E) corre-sponding to a 7-fold decrease in metastatic burden in SS.BN3IL2Rg lungs (41 � 16 RLU/S/mg; P < 0.05) compared withSSIL2Rg lungs (300 � 161 RLU/S/mg; Fig. 3F). Moreover, H&Estaining revealed histologically detectablemetastatic lesions in80% of SSIL2Rg lungs, but only 20% of SS.BN3IL2Rg lungs exam-ined (data not shown). Collectively, these data suggest thatgenetic variants on BN chromosome 3 increase tumor BVD,however, with a paradoxical decrease in hematogenousmetas-tasis that is potentially due to decreased vascular invasion.In most cases, BVD is positively correlated with tumor

growth and hematogenous metastasis (33), prompting us tovalidate whether the paradoxical angiogenesis phenotype seenin SS.BN3IL2Rg (Fig. 3B) is present in the unmodified parental SSand SS.BN3 consomic strains using two independent methods:a Matrigel plug angiogenesis assay (34) and reanalysis of theDMBA-inducible mammary tumors from Adamovic and col-

leagues (16). For the Matrigel plug angiogenesis assay, femaleSS and SS.BN3 consomic rats were implanted with 500 mL ofMatrigel supplemented with 500 ng/mL of recombinant ratVEGF164. At 5 days postimplantation, Matrigel plugs wereexcised and quantified for CD31þ blood vessels across theentire cross section of the plug. Comparedwith SS plugs (14� 3vessels/mm2; n¼ 9 rats), the BVD in SS.BN3 consomic rats wasapproximately 2-fold higher (23� 4 vessels/mm2; P < 0.05; n¼19 rats; Fig. 4A and B). Likewise, BVD of SS.BN3 mammarytumors (n ¼ 9) from the DMBA-inducible model showed anapproximately 2-fold (P < 0.01) increase relative to SS tumors(n¼ 9) from the same study (Fig. 4C and D; ref. 16), collectivelydemonstrating that SS.BN3 has increased angiogenic potential(i.e., the ability to form new blood vessels), despite havingdecreased tumor growth, vascular invasion, and hematoge-nous metastasis (Figs. 2 and 3).

Tumor lymphangiogenesis and lymphogenousmetastasis

To assess LVD, 231Lucþ tumors that were orthotopicallyimplanted in MFPs of SSIL2Rg and SS.BN3IL2Rg rats were excisedat 24 days postimplantation and stainedwith antibodies againstthe lymphatic vesselmarker, LYVE-1 (Fig. 4A; ref. 35). In contrastto tumor BVD, we observed no difference in LVD betweenSSIL2Rg (41 � 7 vessels per field; n ¼ 5 rats) and SS.BN3IL2Rg

Figure3. Characterizationof tumor-associatedblood vessels andhematogenousmetastasis in 231Lucþ tumors implanted inSS.BN3IL2Rg andSSIL2Rg rats at 24days postimplantation. A, visualization of tumor-associated blood vessels using anti-CD31 staining of 231Lucþ tumors implanted in SS.BN3IL2Rg and SSIL2Rg

rats. Images were acquired at�100magnification. Scale bar, 100 mm. B, mean blood vessel density in 231Lucþ tumors implanted in SS.BN3IL2Rg and SSIL2Rg

rats was calculated from three images per tumor (n ¼ 5–6 tumors per strain) acquired at �100 magnification. Data, mean vascular density per �100 field �SEM. �, P < 0.05, as determined by the unpaired Student t test. C, percentage of open lumens in 231Lucþ tumors implanted in SS.BN3IL2Rg and SSIL2Rg wascalculated from three images per tumor (n ¼ 5–6 per strain) acquired at �100 magnification. Data, the percentage of CD31þ vessels with open lumensper�100 field�SEM. �,P < 0.05, as determined by the unpairedStudent t test. D, percentage of CD31þ vessels invaded by 231Lucþ cells in tumors implantedin SS.BN3IL2Rg and SSIL2Rg was calculated from three images per tumor (n¼ 5–6 tumors per strain) acquired at�100 magnification. Data, the percentage ofCD31þ vessels with open lumens per �100 field � SEM. ��, P < 0.001, as determined by the unpaired Student t test. E, representative luminescentimaging of lungs from tumor-bearing SSIL2Rg and SS.BN3IL2Rg rats. F, hematogenous metastatic burden was measured by luciferase activity normalized tototal milligrams of protein in lung lysates from SSIL2Rg (n¼ 12) and SS.BN3IL2Rg (n¼ 16) rats. Metastatic burden of individual rats is represented by the dots,and the black bars indicate the average metastatic burden per strain. �, P < 0.05, as determined by the unpaired Student t test.






(41 � 5 vessels per field; n ¼ 6 rats) tumors (Fig. 4B), whichcoincided with no differences in metastatic burden in theaxillary lymph nodes of either strain (Fig. 4C). Both strainsshowed a high incidence of lymphatic vessel invasion by231Lucþ cells in 80% (SSIL2Rg) and 67% (SS.BN3IL2Rg) of tumorsexamined (data not shown), indicating a similar propensity of231Lucþ tumors toundergo lymphogenousmetastasis inboth ratstrains. Because no differences were detected in LVD, lymphaticinvasion, and lymph node metastasis (Fig. 4), our data suggestthat the causative variants on BN chromosome 3 are vascularcell-type specific.

Sequence analysis of rat chromosome 3Our data demonstrate that genetic elements on BN chro-

mosome 3 affect tumor growth, vascularity, vascular invasion,and metastasis (Figs. 2–5). Overlapping cancer QTLs on ratchromosome 3 were previously reported in independent F2crosses using cancer resistant (BN/SsNHsd and COP/CrCrl)and cancer susceptible (F344/NHsd and ACI/SegHsd) strains,including the Epdm3 QTL (chr3:50.5-88.6 Mb, BN/SsNHsdx F344/NHsd; ref. 17), the Emca5 QTL (chr3:41.0-171.0 Mb,

BN/SsNHsd x ACI/SegHsd; ref. 18), and the Ept2 QTL (chr3:44.6-110.2Mb,ACI/SegHsd xCOP/CrCrl; ref. 19; Fig. 6). Analysis of theF344.BN3 congenic rat (F344.BN-Edpm3BN) showed decreasedtumor susceptibility and increased density of malformed tumorblood vessels (32), which closely resembles the SS.BN3IL2Rg

consomic phenotypes (Figs. 2 and 3). As the protective allelesin the F344.BN-Edpm3BN QTL (17, 32) and the SS.BN3IL2Rg

consomic are BN-derived, the causative alternative alleles inthis region are likely shared by SS and F344. Because thecausative alleles in the Emca5 QTL (18) and Ept2 QTL (19) areACI-derived, the protective alternative alleles in this region arelikely shared by BN and Cop. Combining all data, we extendedthe cosegregating sequence analysis to BN and Cop (protective)versus SS, F344, and ACI (susceptible).

We analyzed the distribution of alleles on rat chromosome 3using WGS of BN/NHsdMcwi, ACI/Eur, Cop/Crl, F344/N, andSS/JrHsdMcwi that was recently generated in collaborationwith Atanur and colleagues (27) (Fig. 6A). To identify cose-gregating and unique genomic regions that could influencecancer susceptibility, we plotted the total unique and commonvariants per Mb that cosegregated between two groups: BN/NHsdMcwi and Cop/Crl (tumor-resistant) versus F344/NHsd,

Figure 4. Increased angiogenic potential in SS.BN3 consomic rats. A,density of CD31þ blood vessels inMatrigel plugs implanted in SS (n¼ 20)and SS.BN3 (n¼ 10) consomic rats. B,meanCD31þblood vessel densityin Matrigel plugs implanted in SS and SS.BN3 consomic rats wascalculated from the entire Matrigel plug cross-section. Data, the meanvascular density per mm2 � SEM. �, P < 0.05, as determined by theunpaired Student t test. C, density of SH2B3þ blood vessels in DMBA-induced tumors of SS (n¼ 9) and SS.BN3 (n¼ 9) consomic rats. D, meandensity of SH2B3þ blood vessels was calculated from �400 images ofthe entire tumor cross-section that were acquired using an automatedmicroscope. Data, the mean vascular density per area � SEM.��, P < 0.01, as determined by the unpaired Student t test.

Figure 5. Characterization of tumor-associated lymphatic vessels andlymphogenous metastasis in 231Lucþ tumors implanted in SS.BN3IL2Rg

and SSIL2Rg rats at 24 days postimplantation. A, visualization of tumor-associated lymphatic vessels using anti-LYVE-1 staining of 231Lucþ

tumors implanted in SS.BN3IL2Rg and SSIL2Rg rats. Scale bar, 100 mm.B, mean lymphatic vessel density in 231Lucþ tumors implanted in SS.BN3IL2Rg and SSIL2Rg rats was calculated from three images per tumor(n ¼ 5–6 per strain) acquired at �100 magnification. Data, the meanvascular density per �100 field � SEM. C, lymphogenous metastaticburdenwasmeasured by luciferase activity normalized to totalmilligramsof protein in axillary lymph node lysates from SS.BN3IL2Rg (n ¼ 5) andSSIL2Rg (n¼ 8) rats. Metastatic burden of individual rats is represented bythe dots, and the black bars indicate the average metastatic burden perstrain. For B and C, statistical analysis of data was performed by theunpaired Student t test.

Flister et al.





ACI/SegHsd, and SS/JrHsdMcwi (tumor-susceptible; Fig. 6B).The overlapping QTL regions (chr3:50.5-88.0 Mb) were signif-icantly enriched by 30-fold (P < 0.001) for cosegregating allelescompared to the average of the rest of chromosome 3 (Fig. 6A–C), indicating that the genetic mechanisms underlying thesesimilar QTLs are likely shared in these regions. Combined,these data suggest that common heritable elements are mostlikely enriched in the overlapping cancer QTLs (16–19).Using the cosegregating variant analysis we preliminarily

reduced the list of candidate variants from 455,272 totalvariants on rat chromosome 3 to 15,031 variants (3.3% of totalchromosome 3 variants) in the cosegregating region (chr3:50.5-88.0 Mb) that overlapped with the three cancer QTLs (17–19).Of the 15,031 cosegregating variants, 46 lie within codingregions of conserved genes, 10 are expected to cause nonsy-nonymous amino acid changes, and 2 of these were predictedto be damaging by Provean (http://provean.jcvi.org/seq_sub-mit.php; Table 1 and Supplementary Table S3).

Comparative genomicsOur data suggest that cosegregating genetic elements on

rat chromosome 3 affect tumor growth, angiogenesis, vas-

cular invasion, and metastasis (Figs. 2–5A). The 37.5-Mbregion (chr3:50.5–88.0 Mb) of overlapping cancer QTLscontains 184 conserved and validated genes (SupplementaryTable S2) and is syntenic to two regions on human chromo-somes 2 and 11 (Fig. 6D) and one region on mouse chro-mosome 2 (data not show). A total of four genes [METAP1D(5), CDCA7 (5), PTPRJ (36), and CD44 (37–40)] in syntenicregions of the human genome have been associated withbreast cancer risk (Fig. 6D), whereas no mouse studies havereported associations for breast cancer in the syntenicmouse region. Of the human risk genes, METAP1D andCDCA7 were associated with breast cancer incidence bygenome-wide association study (5). CD44 has been associ-ated with multiple aspects of human breast cancer risk:incidence (39), age of onset (37, 40), angiogenesis (41),metastasis (42), and survival (39, 43, 44). PTPRJ mutationswere associated with breast cancer risk by loss of hetero-zygosity analysis (36) and PTPRJ has also been implicated inangiogenesis (45). Strikingly, PTPRJ mutation caused disor-ganized vascular plexus formation in mouse embryos (46),which was marked by increased density of poorly formedblood vessels that are reminiscent of the vascular phenotypes

Figure 6. Sequence analysis of cosegregating alleles and comparative mapping of rat chromosome 3. A, Venn diagram of sequence comparison of the entirerat chromosome 3 between variants shared by the cancer-resistant BN/NHsdMcwi and Cop/Crl strains versus the cancer-susceptible strains ACI/Eur, F344/N, and SS/JrHsdMcwi. B, density mapping of variants shared by the cancer-resistant BN/NHsdMcwi and Cop/Crl strains versus the cancer-susceptiblestrains ACI/Eur, F344/N, and SS/JrHsdMcwi. Data, the number of variants per 1-Mb bin. Note the highest density of cosegregating variants overlap with threebreast cancer QTLs (labeled brackets). C, Venn diagram of sequence comparison of the shared QTL region of chromosome 3 (50.5–88.0 Mb) betweenvariants shared by the cancer-resistant BN/NHsdMcwi and Cop/Crl strains versus the cancer-susceptible strains ACI/Eur, F344/N, and SS/JrHsdMcwi.D, schematic representation of the overlapping cancer QTLs on rat chromosome 3 and the orthologous regions in human. The overlapping human breastcancer loci are indicated.






associatedwith the SS.BN3IL2Rg consomic (Fig. 3) and F344.BN-Edpm3BN congenic (32) models.

DiscussionThe heritability of breast cancer is well established (1–3), yet

the majority of causative variants remain unknown (5). Of thefew known variants, most are considered tumor cell-autono-mous (47–49), with far less emphasis placed on germlinevariants that might affect breast cancer progression throughthe tumor microenvironment. We developed an experimentalmodel that specifically focused on mapping genetic variantsunderlying differences in the tumor microenvironment. CXMutilizes genetically "fixed" tumor cells that are xenografted intorat strains with different genetic backgrounds, so that anyobserved differences in tumor progression (e.g., growth, vas-culogenesis, and metastasis) would be attributable to thetumor microenvironment, including interaction between thetumor and host. From a clinical perspective, these data showthat germline variants in the tumor microenvironment affectbreast cancer risk and should be taken into account in futurestudies. As a technical advancement, this work demonstratesthat CXM can be used to directly map germline variants in thetumor microenvironment for the first time.

Using CXM, we demonstrated that genetic variants on BNchromosome 3 inhibited tumor growth and lung metastasis of231Lucþ breast cancer cells (Figs. 2 and 3), without affectinglymph node metastasis (Fig. 5). Sequence analysis by cross-strain comparisons and comparative analysis prioritized a listof 6 candidate genes for future study (Fig. 6; Table 1). These 6candidate genes were prioritized using functional predictionsof nonsynonymous variants (Fsip2 and Pramel7) and sharedassociation with human breast cancer [METAP1D (5), CDCA7(5), PTPRJ (36), and CD44 (37–40)]. Within the region, it isalso possible that other nonsynonymous variants and inter-genic variants have functional consequences beyond thelimitations of prediction software. Thus, validating thesecandidates genes will likely require further narrowing of the

QTL by congenics (26, 50, 51), gene expression analysis,and/or directly testing candidates by gene editing (20, 52,53), all of which we have done previously for other diseaseloci (see below for description).

Breast cancer risk QTL on rat chromosome 3Previously, we (16) and others (17–19) identified overlapping

cancer QTLon rat chromosome 3 (50.5–88.0Mb). A congenic ofthe Edpm3 QTL (F344.BN-Edpm3BN) indicated that the pro-tective cancer QTL is likely due to malformed immature bloodvasculature, despite the tumors having a higher BVD (32).Similarly, we observed significant tumor growth inhibition inthe SS.BN3IL2Rg consomic comparedwith SSIL2Rg (Fig. 2); with aparadoxical increase in tumor-associated blood vessels (Fig. 3Aand B), suggesting that the vascular phenotypes underlying therat chromosome 3 QTL are likely derived from the tumormicroenvironment. Tumor metastasis in the previous rat QTLstudies was not reported (17–19, 32), as most inducible ratmammary tumors are poorlymetastatic. By CXM,wewere ableto show that variants on BN chromosome 3 also affect vascularinvasion and distal metastasis of 231Lucþ cells (Figs. 3 and 5).Despite the increased BVD in SS.BN3IL2Rg , both vascular inva-sion of tumor cells and hematogenous metastatic burden weresignificantly lower in SS.BN3IL2Rg (Fig. 3C–E), suggesting apotential mechanism that attenuates metastasis by limitinginvasion or transport of tumor cells. Collectively, CXM extend-ed the previous findings (17–19) by demonstrating that (i)genetic variants in the rat chromosome 3 QTL likely actthrough the tumormicroenvironment; (ii) the geneticmechan-isms are blood vessel specific; and (iii) tumor growth andhematogenous metastasis are also attenuated by the geneticvariants.

Blood vasculature is essential for tumor growth and meta-static progression (54). Our data suggest that germline variantsaffecting tumor-associated blood vessels can significantlyaffect cancer risk (Figs. 2 and 3). Vascular defects also decreaserisk of solid tumors in Down syndrome patients with trisomy21, due to increased gene dosage of DSCR1 (55). The important

Table 1. Cosegregating nonsynonymous variants in the overlapping cancer QTL on rat chromosome 3

Genesymbol Gene ID

Referencenucleotide

Variantnucleotide Position aa

Proveanprediction

Lrp2 68407 G A 51,642,105 S1443L �2.026Fsip2 1593013 A T 66,068,203 E4965D �2.625Pramel7 1565990 T A 71,166,422 D435V �5.107Pramel6 1565946 T G 71,183,838 F462C 1.346Pacsin3 1307327 G T 75,614,920 G13V �0.473Rag2 1305588 T C 86,771,280 I387T 0.584Rag1 619790 A G 86,785,697 V503A 1.105Rag1 619790 G C 86,786,211 Q334E 0.357Pamr1 1308745 A C 87,845,730 E299A 0.105Slc1a2 3697 A G 87,988,333 I515V �0.020

NOTE: Provean prediction of�2.5 or higher (emboldened) indicates that the amino acid change is likely damaging to protein function.Abbreviation: aa, amino acid.

Flister et al.





distinction is that Down syndrome is a non-mendelian geneticdisease, whereas our findings suggest that common germlinevariants could have similar effects in amuch broader spectrumof patients. Our data also suggest that "looks can be deceiving",i.e., metastatic potential and tumor growth are frequentlycorrelated with vascular density (33), but based on our find-ings, thismight not always be the case. Instead,we propose thatvariants affecting vascular function and maturity will similarlyaffect breast cancer progression, whichmight be overlooked asa prognostic indicator by relying on vascular density only.Finally, from a translational viewpoint, the findings of thisstudy indicate that some germline variants could potentiallyfunction as "genetic switches" between hematogenous andlymphogenous metastasis. Thus, a better understanding ofgermline variants that affect tumor metastasis will likely yieldbetter patient stratification for vascular-specific adjuvant ther-apies [e.g., bevacizumab (56), TKi (57), etc.].

Using CXM and similar strategies to localize cancer riskvariantsFrom a technical standpoint, we posit that CXM can be

tailored to further elucidate causative variants using additionalmapping strategies (e.g., F2 intercross, congenic, gene editing,etc.) or to test interactions with other clinically relevantsubsets of breast cancer (e.g., luminal, lobular, inflammatory,etc.). Our rationale for first using a consomic was (i) to developthe general technique and (ii) to localize the phenotype to aspecific chromosome before proceeding further. Using CXMand comparative mapping, we narrowed the list of potentialcandidate variants in the rat chromosome 3 QTL by 96.7%.Now, it would be foreseeable to physically narrow the QTL,with the ultimate goal of localizing the causative variant.Because TALENs and ZFNs do not require specific rat strains(we have now gene edited on >10 backgrounds; http://rgd.mcw.edu/wg/physgenknockouts), it would be possible to gen-erate SCID rats (ormice) on any background. For example, onecould reduce the SS.BN3 cancer QTL by selectively backcross-ing SS.BN3IL2Rg to SSIL2Rg until the region of BN chromosome 3has been sufficiently narrowed to resolve the causative variant,followed by expression analysis and TALEN- or ZFN-mediatedgene editing of candidate genes on the minimal SS.BN3IL2Rg

congenic background, an approach that we have used previ-ously for other disease loci (53). Finally, although we developedCXM for the rat, we have also demonstrated the utility ofTALEN- or ZFN-mediated gene editing in the mouse (58) andtherefore, CXM could be used in mice and potentially othermodel organisms.

Experimental considerations of the CXM strategyThere are several limitations of cancer xenograft models

that should be considered when interpreting CXM data. First,like all xenograft models, CXM utilizes SCID rodents that lackfully intact immune systems and thus genetic variants thataffect tumor immunity might not be accessible. Second, xeno-transplantation of human cancer cells into rodent back-grounds might not recapitulate species-specific mechanisms,but rather only mechanisms conserved across species. Finally,some germline variants likely affect bothmalignant tumor cells

and the nonmalignant stromal cells, whereas CXM would onlydetect effects on the nonmalignant stroma. Despite theselimitations, there are also considerable advantages of CXM,including the ability to use etiologically defined and clinicallyrelevant human cell lines that are transgenically tagged forbetter tracking in vivo.

PerspectiveCXM is the first model for mapping germline variants in the

tumor microenvironment. To do so, CXM combines genomeediting (via TALENs), a genetic mapping tool (the consomic),and a widely used model of human breast cancer (the tumorxenograft). As proof-of-concept, we demonstrated that CXMcan detect germline variants in an established breast cancerQTL (Figs. 2 and 3; refs. 17–19). Using CXM, we also demon-strated that the causative variants on rat chromosome 3affected the tumor blood vasculature, which potentially dis-rupted tumor growth and metastasis. Follow-up studies usingcell culture systems, localization of candidate gene expression,and tissue-specific gene editing will be required to determinewhether the causative variants function directly or indirectlythrough the blood vasculature. In summary, CXM is a technicaladvance that localizes cancer phenotypes to a single chromo-some and could be used to isolate causative variants/genes bycongenic mapping and/or candidate gene targeting.

Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.

Authors' ContributionsConception and design: M.J. Flister, A.M. Geurts, J. LazarDevelopment of methodology: M.J. Flister, B.T. Endres, A.M. Geurts, S. Ran,D. Carlson, S.C. FahrenkrugAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): M.J. Flister, B.T. Endres, N. Rudemiller, A. Lemke,I. Roy, A.M. Geurts, C. Moreno, P.E. North, M.B. Dwinell, J.D. ShullAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): M.J. Flister, B.T. Endres, N. Rudemiller, A.B. Sarkis,S. Santarriaga, S.-W. Tsaih, J.D. Pons, P.E. North, P.S. LaVioletteWriting, review, and/or revision of the manuscript: M.J. Flister, N. Rude-miller, A.B. Sarkis, C. Moreno, Z. Lazarova, J. Lazar, P.S. LaViolette, M.B. Dwinell,J.D. Shull, H.J. JacobAdministrative, technical, or material support (i.e., reporting or orga-nizing data, constructing databases): C. Moreno, W. TanStudy supervision: J. Lazar, H.J. Jacob

AcknowledgmentsThe authors thank M. Teisch, E. Christenson, C. Duris, M. Tschannen,

J. Wendt-Andrae, M. Grzybowski, S. Kalloway, J. Foeckler, and R. Schilling forexcellent technical support.

Grant SupportThis study was supported by an Interdisciplinary Cancer Research Fellowship

from the Medical College of Wisconsin Cancer Center (M.J. Flister), a FroedtertHospital Foundation Skin Cancer Center Research Award (Z. Lazar), NCIgrant R01CA77876 (J.D. Shull), and NHLBI grants 5RC2HL101681 andR24HL11447401A1 (H.J. Jacob). M.J. Flister is also supported by an NHLBItraining grant (5T32HL007792). A.M. Geurts is supported by an NIH Director'sNew Innovator Award (DP2OD008396).

The costs of publication of this article were defrayed in part by the payment ofpage charges. This article must therefore be hereby marked advertisement inaccordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Received November 12, 2013; revised July 8, 2014; accepted July 9, 2014;published OnlineFirst August 29, 2014.






References1. Peto J, Mack TM. High constant incidence in twins and other relatives

of women with breast cancer. Nat Genet 2000;26:411–4.2. Rosman DS, Kaklamani V, Pasche B. New insights into breast cancer

genetics and impact on patient management. Curr Treat OptionsOncol 2007;8:61–73.

3. Pharoah PD, Antoniou A, Bobrow M, Zimmern RL, Easton DF, PonderBA. Polygenic susceptibility to breast cancer and implications forprevention. Nat Genet 2002;31:33–6.

4. Hartman M, Lindstrom L, Dickman PW, Adami HO, Hall P, Czene K. Isbreast cancer prognosis inherited? Breast Cancer Res 2007;9:R39.

5. MichailidouK,Hall P, Gonzalez-Neira A,GhoussainiM,Dennis J,MilneRL, et al. Large-scale genotyping identifies 41 new loci associatedwithbreast cancer risk. Nat Genet 2013;45:353–61, 361e351–352.

6. Rafiq S, Tapper W, Collins A, Khan S, Politopoulos I, Gerty S, et al.Identification of inherited genetic variations influencing prognosis inearly-onset breast cancer. Cancer Res 2013;73:1883–91.

7. Park YG, Zhao X, Lesueur F, Lowy DR, Lancaster M, Pharoah P, et al.Sipa1 is a candidate for underlying the metastasis efficiency modifierlocus Mtes1. Nat Genet 2005;37:1055–62.

8. Shu XO, Long J, LuW, Li C, ChenWY, Delahanty R, et al. Novel geneticmarkers of breast cancer survival identified by a genome-wide asso-ciation study. Cancer Res 2012;72:1182–9.

9. Minn AJ, Gupta GP, Siegel PM, Bos PD, Shu W, Giri DD, et al. Genesthat mediate breast cancer metastasis to lung. Nature 2005;436:518–24.

10. Gupta GP, Nguyen DX, Chiang AC, Bos PD, Kim JY, Nadal C, et al.Mediators of vascular remodelling co-opted for sequential steps inlung metastasis. Nature 2007;446:765–70.

11. Maae E, Andersen RF, Steffensen KD, Jakobsen EH, Brandslund I,Sorensen FB, et al. Prognostic impact of VEGFA germline polymorph-isms in patients with HER2-positive primary breast cancer. AnticancerRes 2012;32:3619–27.

12. Smith KC, Bateman AC, Fussell HM, Howell WM. Cytokine genepolymorphisms and breast cancer susceptibility and prognosis. EurJ Immunogenet 2004;31:167–73.

13. Cowley AW Jr, Roman RJ, Jacob HJ. Application of chromosomalsubstitution techniques in gene-function discovery. J Physiol 2004;554:46–55.

14. Carlson DF, Fahrenkrug SC, Hackett PB. Targeting DNA with fingersand TALENs. Mol Ther Nucleic Acids 2012;1:e3.

15. Mashimo T, Takizawa A, Voigt B, Yoshimi K, Hiai H, Kuramoto T, et al.Generation of knockout rats with X-linked severe combined immuno-deficiency (X-SCID) using zinc-finger nucleases. PLoS ONE 2010;5:e8870.

16. Adamovic T, McAllister D, Wang T, Adamovic D, Rowe JJ, Moreno C,et al. Identification of novel carcinogen-mediated mammary tumorsusceptibility loci in the rat using the chromosome substitution tech-nique. Genes Chromosomes Cancer 2010;49:1035–45.

17. Wendell DL, Gorski J. Quantitative trait loci for estrogen-dependentpituitary tumor growth in the rat. Mamm Genome 1997;8:823–9.

18. Schaffer BS, Lachel CM, Pennington KL, Murrin CR, Strecker TE,Tochacek M, Genetic bases of estrogen-induced tumorigenesis in therat: mapping of loci controlling susceptibility to mammary cancer in aBrown Norway x ACI intercross. Cancer Res 2006;66:7793–800.

19. Kurz SG, Hansen KK, McLaughlin MT, Shivaswamy V, Schaffer BS,Gould KA, et al. Tissue-specific actions of the Ept1, Ept2, Ept6, andEpt9 genetic determinants of responsiveness to estrogens in thefemale rat. Endocrinology 2008;149:3850–9.

20. Geurts AM, Cost GJ, Freyvert Y, Zeitler B, Miller JC, Choi VM, et al.Knockout rats via embryo microinjection of zinc-finger nucleases.Science 2009;325:433.

21. Volk LD, Flister MJ, Bivens CM, Stutzman A, Desai N, Trieu V, et al.Nab-paclitaxel efficacy in the orthotopic model of human breastcancer is significantly enhanced by concurrent anti-vascular endothe-lial growth factor A therapy. Neoplasia 2008;10:613–23.

22. Faraji F, Hu Y, Wu G, Goldberger NE, Walker RC, Zhang J, et al. Anintegrated systems genetics screen reveals the transcriptional struc-ture of inherited predisposition to metastatic disease. Genome Res2014;24:227–40.

23. Whitehurst B, Flister MJ, Bagaitkar J, Volk L, Bivens CM, Pickett B,et al. Anti-VEGF-A therapy reduces lymphatic vessel density andexpression of VEGFR-3 in an orthotopic breast tumor model. Int JCancer 2007;121:2181–91.

24. Hall KL, Volk-Draper LD, Flister MJ, Ran S. Newmodel of macrophageacquisition of the lymphatic endothelial phenotype. PLoSONE2012;7:e31794.

25. Mattson DL, Lund H, Guo C, Rudemiller N, Geurts AM, Jacob H.Genetic mutation of recombination activating gene 1 in Dahl salt-sensitive rats attenuates hypertension and renal damage. AmJPhysiolRegul Integr Comp Physiol 2013;304:R407–14.

26. Flister MJ, Prisco SZ, Sarkis AB, O'Meara CC, Hoffman M, Wendt-Andrae J, et al. Identification of hypertension susceptibility loci on ratchromosome 12. Hypertension 2012;60:942–8.

27. Atanur SS, Diaz AG, Maratou K, Sarkis A, Rotival M, Game L, et al.Genome sequencing reveals loci under artificial selection that underliedisease phenotypes in the laboratory rat. Cell 2013;154:691–703.

28. Gibbs RA, Weinstock GM, Metzker ML, Muzny DM, Sodergren EJ,Scherer S, et al. Genome sequence of the Brown Norway rat yieldsinsights into mammalian evolution. Nature 2004;428:493–521.

29. Lifsted T, Le Voyer T, Williams M, Muller W, Klein-Szanto A, BuetowKH, et al. Identification of inbred mouse strains harboring geneticmodifiers ofmammary tumor age of onset andmetastatic progression.Int J Cancer 1998;77:640–4.

30. Ohbo K, Suda T, HashiyamaM,Mantani A, IkebeM,Miyakawa K, et al.Modulation of hematopoiesis in mice with a truncated mutant of theinterleukin-2 receptor gamma chain. Blood 1996;87:956–67.

31. Flister MJ, Wilber A, Hall KL, Iwata C, Miyazono K, Nisato RE, et al.Inflammation induces lymphangiogenesis through up-regulation ofVEGFR-3 mediated by NF-kappaB and Prox1. Blood 2010;115:418–29.

32. Pandey J, Wendell DL. Angiogenesis and capillary maturation pheno-types associated with the Edpm3 locus on rat chromosome 3. MammGenome 2006;17:49–57.

33. Ran S, Volk L, Hall K, Flister MJ. Lymphangiogenesis and lymphaticmetastasis in breast cancer. Pathophysiology 2010;17:229–51.

34. Auerbach R, Lewis R, Shinners B, Kubai L, Akhtar N. Angiogenesisassays: a critical overview. Clin Chem 2003;49:32–40.

35. Flister MJ, Volk LD, Ran S. Characterization of Prox1 and VEGFR-3expression and lymphatic phenotype in normal organs of mice lackingp50 subunit of NF-kappaB. Microcirculation 2011;18:85–101.

36. RuivenkampCA, vanWezel T, ZanonC,StassenAP, VlcekC,CsikosT,et al. Ptprj is a candidate for the mouse colon-cancer susceptibilitylocus Scc1 and is frequently deleted in human cancers. Nat Genet2002;31:295–300.

37. Zhou J, Nagarkatti PS, ZhongY, Zhang J, Nagarkatti M. Implications ofsingle nucleotide polymorphisms in CD44 exon 2 for risk of breastcancer. Eur J Cancer Prev 2011;20:396–402.

38. Zhou J, Nagarkatti PS, Zhong Y, Creek K, Zhang J, Nagarkatti M.UniqueSNP inCD44 intron 1and its role in breast cancer development.Anticancer Res 2010;30:1263–72.

39. Jiang L, Deng J, Zhu X, Zheng J, You Y, Li N, et al. CD44 rs13347 C>Tpolymorphism predicts breast cancer risk and prognosis in Chinesepopulations. Breast Cancer Res 2012;14:R105.

40. Tulsyan S, Agarwal G, Lal P, Agrawal S, Mittal RD,Mittal B. CD44 genepolymorphisms in breast cancer risk and prognosis: a study in NorthIndian population. PLoS ONE 2013;8:e71073.

41. Cao G, Savani RC, Fehrenbach M, Lyons C, Zhang L, Coukos G, et al.Involvement of endothelial CD44 during in vivo angiogenesis. Am JPathol 2006;169:325–36.

42. Lopez JI, Camenisch TD, Stevens MV, Sands BJ, McDonald J,Schroeder JA. CD44 attenuates metastatic invasion during breastcancer progression. Cancer Res 2005;65:6755–63.

43. Diaz LK, Zhou X,Wright ET, Cristofanilli M, Smith T, YangY, et al. CD44expression is associated with increased survival in node-negativeinvasive breast carcinoma. Clin Cancer Res 2005;11:3309–14.

44. Berner HS, Suo Z, Risberg B, Villman K, Karlsson MG, Nesland JM.Clinicopathological associations of CD44 mRNA and protein expres-sion in primary breast carcinomas. Histopathology 2003;42:546–54.

Flister et al.





45. Takahashi T, Takahashi K,Mernaugh RL, Tsuboi N, Liu H, Daniel TO. Amonoclonal antibody against CD148, a receptor-like tyrosine phos-phatase, inhibits endothelial-cell growth and angiogenesis. Blood2006;108:1234–42.

46. Takahashi T, Takahashi K, St John PL, Fleming PA, Tomemori T,Watanabe T, et al. A mutant receptor tyrosine phosphatase, CD148,causesdefects invasculardevelopment.MolCell Biol 2003;23:1817–31.

47. Nguyen DX, Massague J. Genetic determinants of cancer metastasis.Nat Rev Genet 2007;8:341–52.

48. Hunter K. Host genetics influence tumour metastasis. Nat Rev Cancer2006;6:141–6.

49. Bodenstine TM, Welch DR. Metastasis suppressors and the tumormicroenvironment. Cancer Microenviron 2008;1:1–11.

50. Flister MJ, Hoffman MJ, Reddy P, Jacob HJ, Moreno C. Congenicmapping and sequence analysis of the Renin locus. Hypertension2013;61:850–6.

51. Hoffman MJ, Flister MJ, Nunez L, Xiao B, Greene AS, Jacob HJ, et al.Female-specific hypertension Loci on rat chromosome 13. Hyperten-sion 2013;62:557–63.

52. Flister MJ, Tsaih SW,O'Meara CC, Endres B, HoffmanMJ, Geurts AM,et al. Identifying multiple causative genes at a single GWAS locus.Genome Res 2013;23:1996–2002.

53. Rangel-Filho A, Lazar J, Moreno C, Geurts A, Jacob HJ. Rab38modulates proteinuria in model of hypertension-associated renal dis-ease. J Am Soc Nephrol 2013;24:283–92.

54. Hanahan D, Folkman J. Patterns and emerging mechanisms of theangiogenic switch during tumorigenesis. Cell 1996;86:353–64.

55. Baek KH, Zaslavsky A, Lynch RC, Britt C, Okada Y, Siarey RJ, et al.Down's syndrome suppression of tumour growth and the role of thecalcineurin inhibitor DSCR1. Nature 2009;459:1126–30.

56. Ferrara N. Vascular endothelial growth factor: basic science andclinical progress. Endocr Rev 2004;25:581–611.

57. Christensen JG. A preclinical review of sunitinib, a multitargetedreceptor tyrosine kinase inhibitor with anti-angiogenic and antitumouractivities. Ann Oncol 2007;18:x3–10.

58. Chen YG, Forsberg MH, Khaja S, Ciecko AE, Hessner MJ, Geurts AM.Gene targeting in NOD mouse embryos using zinc-finger nucleases.Diabetes 2013;63:68–74.






2014;74:6419-6429. Published OnlineFirst August 29, 2014.Cancer Res Michael J. Flister, Bradley T. Endres, Nathan Rudemiller, et al. MicroenvironmentCXM: A New Tool for Mapping Breast Cancer Risk in the Tumor

Updated version

10.1158/0008-5472.CAN-13-3212doi:

Access the most recent version of this article at:

Material

Supplementary

http://cancerres.aacrjournals.org/content/suppl/2014/08/29/0008-5472.CAN-13-3212.DC1

Access the most recent supplemental material at:

Cited articles

http://cancerres.aacrjournals.org/content/74/22/6419.full#ref-list-1

This article cites 58 articles, 16 of which you can access for free at:

Citing articles

http://cancerres.aacrjournals.org/content/74/22/6419.full#related-urls

This article has been cited by 1 HighWire-hosted articles. Access the articles at:

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

Subscriptions

Reprints and

[email protected]

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

Permissions

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

.http://cancerres.aacrjournals.org/content/74/22/6419To request permission to re-use all or part of this article, use this link



http://cancerres.aacrjournals.org/lookup/doi/10.1158/0008-5472.CAN-13-3212

http://cancerres.aacrjournals.org/content/suppl/2014/08/29/0008-5472.CAN-13-3212.DC1

http://cancerres.aacrjournals.org/content/74/22/6419.full#ref-list-1

http://cancerres.aacrjournals.org/content/74/22/6419.full#related-urls

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

mailto:[email protected]

http://cancerres.aacrjournals.org/content/74/22/6419


cxm: a new tool for mapping breast cancer risk in the ... · cxm: a new tool for mapping breast...

Documents