Introduction of oncogenes into mammary glandsin vivo with an avian retroviral vector initiatesand promotes carcinogenesis in mouse modelsZhijun Du*, Katrina Podsypanina†, Shixia Huang*, Amanda McGrath*, Michael J. Toneff*,Ekaterina Bogoslovskaia*, Xiaomei Zhang*, Ricardo C. Moraes*, Michele Fluck‡,D. Craig Allred*, Michael T. Lewis*§, Harold E. Varmus†¶, and Yi Li*§¶

*Breast Center and §Department of Molecular and Cellular Biology, Baylor College of Medicine, Houston, TX 77030; †Program in Cancer Biologyand Genetics, Memorial Sloan-Kettering Cancer Center, New York, NY 10021; and ‡Department of Microbiology, Michigan State University,East Lansing, MI 48824

Contributed by Harold E. Varmus, September 29, 2006 (sent for review September 7, 2006)

We have adapted the avian leukosis virus RCAS (replication-competent avian sarcoma-leukosis virus LTR splice acceptor)-me-diated somatic gene transfer technique to introduce oncogenesinto mammary cells in mice transgenic for the avian subgroup Areceptor gene, tva, under control of the mouse mammary tumorvirus (MMTV) promoter. Intraductal instillation of an RCAS vectorcarrying the polyoma middle T antigen (PyMT) gene (RCAS-PyMT)induced multiple, oligoclonal tumors within 3 weeks in infectedmammary glands of MMTV-tva transgenic mice. The rapid appear-ance of these tumors from a relatively small pool of infected cells(estimated to be �2 � 103 cells per gland by infection with RCAScarrying a GFP gene; RCAS-GFP) was accompanied by a highfraction of cells positive for Ki67, Cyclin D1, and c-Myc, implyingstrong proliferation competence. Furthermore, the tumors dis-played greater cellular heterogeneity than did tumors arising inMMTV-PyMT mice, suggesting that RCAS-PyMT transforms a rela-tively immature cell type. Infection of mice transgenic for bothMMTV-Wnt-1 and MMTV-tva with RCAS virus carrying an activatedNeu oncogene dramatically enhanced tumor formation over whatis observed in uninfected bitransgenic animals. We conclude thatinfection of mammary glands with retrovirus vectors is an efficientmeans to screen candidate oncogenes for their capacity to initiateor promote mammary carcinogenesis in the mouse.

breast cancer � ErbB2 � tumor virus A � Wnt � polyoma middle T antigen

Genetic changes suspected to be involved in the induction orprogression of breast cancer are conventionally tested by

the generation of transgenic and knockout mice bearing thesechanges (1). However, it is both time-consuming and expensiveto create such germ-line mutants and to cross mutant lines toidentify collaborating mutations. Germ-line models are alsolimited for understanding the initiation and progression of breastcancer because most human breast cancers evolve in a field ofsurrounding normal breast cells, and germ-line alterations ofcancer genes can impair development of the mammary gland (2).

To circumvent some of these difficulties, we have been de-veloping an alternative strategy, using avian retroviral vectors tointroduce potentially oncogenic genetic changes into a subset ofsomatic cells from mouse target organs in vivo or in vitro (3).RCASBP(A) (replication-competent avian sarcoma-leukosis vi-rus LTR splice acceptor, Bryan-Rous sarcoma virus polymerase,subgroup A), referred to as RCAS or RCAS(A) in this article,is a viral expression vector modified from replication-competentsubgroup A avian leukosis virus. High titer stocks of this viruscan be propagated in chicken fibroblasts (3, 4), but RCAS(A)does not infect mammalian cells because mammals do notencode the virus receptor tumor virus A (TVA). However,transgenic expression of avian DNA containing tva rendersotherwise resistant mouse cells susceptible to virus infection.Thus, RCAS(A) vectors may be used to introduce and express

exogenous genes (such as oncogenes) in mouse cells pro-grammed to produce TVA. However, RCAS virus is not pro-duced in the infected mouse cells; hence, the virus does notspread to other cells.

The RCAS-based gene transfer method was initially describedfor introducing genes to myoblasts (5) and has since been usedto induce cancers of the brain (6, 7), ovary (8), vascularendothelium (9), pancreas (10), liver (11), and other cell types(12, 13). In this study, we describe transgenic mice expressing tvaselectively in mammary epithelial cells under the influence of themouse mammary tumor virus (MMTV) LTR. We have usedMMTV-tva mice to determine the efficiency of infection in vivoafter intraductal injection, and we have shown that RCAS(A)vectors carrying oncogenes can induce tumors swiftly, despiteinfection of only a few thousand cells, and can acceleratetumorigenesis initiated by oncogenic transgenes. Thus, RCAS-mediated gene delivery may be useful for testing collections ofgenes for oncogenic potential and collaborative genetic inter-actions in vivo.

ResultsUsing RCAS Vectors to Transfer Genes to Normal Mammary EpithelialCells in Vivo. We generated 10 founder transgenic mice carryingtva under the transcriptional control of the MMTV LTRs (Fig.1A). Using immunohistochemical staining for TVA, we identi-fied two independent transgenic lines (MA and MD) thatexpress tva in the mammary luminal cells in a patchy pattern (Fig.1B), as is commonly observed for other MMTV-regulatedtransgenes (14). As expected, glands from nontransgenic litter-mates did not stain with antiserum specific for TVA (Fig. 1C).Coimmunostaining confirmed that TVA was produced in cellspositive for the epithelial marker keratin 8 (Fig. 1D), but not incells positive for �-smooth muscle actin (�-SMA), a myoepithe-lial marker (Fig. 1E). Only the MA line was used for theremainder of this study.

To document that TVA� mammary epithelial cells are sus-ceptible to infection by RCAS viruses, we isolated cells frommammary glands of MA mice and infected them in culture withRCAS virus carrying the gene encoding GFP (RCAS-GFP).

Author contributions: Z.D., H.E.V., and Y.L. designed research; Z.D., K.P., S.H., A.M., M.J.T.,E.B., X.Z., D.C.A., and Y.L. performed research; R.C.M., M.F., and M.T.L. contributed newreagents�analytic tools; Z.D. and Y.L. analyzed data; and Z.D., H.E.V., and Y.L. wrote thepaper.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.

Abbreviations: RCAS, replication-competent avian sarcoma-leukosis virus LTR splice accep-tor; TVA, tumor virus A; MMTV, mouse mammary tumor virus; PyMT, polyoma middle Tantigen; SMA, smooth muscle actin.

¶To whom correspondence may be addressed. E-mail: liyi@bcm.edu or varmus@mskcc.org.

© 2006 by The National Academy of Sciences of the USA

17396–17401 � PNAS � November 14, 2006 � vol. 103 � no. 46 www.pnas.org�cgi�doi�10.1073�pnas.0608607103

Dow

nloa

ded

by g

uest

on

July

20,

202

1

Three days after infection, we detected strong GFP signals in�10% of the cells, but only in dishes receiving the virus (Fig. 1F)and not in mock-infected dishes (Fig. 1G).

To determine whether TVA� mammary epithelial cells couldalso be infected in vivo, we exposed them to RCAS-GFP viruses(107 units in 10 �l) via an intraductal injection (15) of MA femalemice at 6 weeks of age; at this stage, the animals are in puberty,and mammary epithelial cells are proliferating and thereforesusceptible to infection by onco-retroviruses like RCAS (3).Four days after injection (to ensure adequate time for virus

entry, synthesis, and integration of viral DNA, and expression ofGFP from the provirus), we collected the injected mammaryglands, prepared paraffin sections, and examined them forproduction of GFP by immunohistochemistry, using antibodiesagainst GFP. Staining for GFP was observed in a small numberof ductal cells (Fig. 1H), demonstrating that TVA� mammaryepithelial cells can be infected by RCAS viruses in vivo and invitro. As expected, no staining was found in sections of controlmammary glands from nontransgenic mice that were injected atthe same time with the same virus stock (Fig. 1I), confirming therequirement for the receptor. Using flow cytometry analysis ofsingle cell suspensions made from six pooled mammary glandsfrom each of five infected mice, we also determined that �0.31%� 0.24% of the mammary epithelial cells, or 1,928 � 1,493epithelial cells per gland, were infected by RCAS-GFP in theseexperiments (Fig. 4, which is published as supporting informa-tion on the PNAS web site).

Infection of Mammary Glands of MMTV-tva Transgenics with RCAS-Polyoma Middle T Antigen (PyMT) Induces Hyperplastic Lesions withHigh Levels of Cell Proliferation and Enhanced Expression of Cyclin D1and c-Myc. To test the response of normal mammary epithelialcells to oncogenes delivered by intraductal infection, we injectedRCAS virus carrying the gene encoding PyMT (RCAS-PyMT)via the nipple duct into three glands in each of two MA mice (107

units per gland) at 6 weeks of age. We chose PyMT, a c-Src-binding membrane-bound viral protein that activates multiplesignaling pathways including those mediated by Shc and phos-phatidylinositol 3-kinase (16), because its strong oncogenicpotential increased the likelihood that tumors would result frominfection of relatively small numbers of cells. Indeed, even at 7days after infection, intraductal lesions were readily detectablein H&E-stained sections (Fig. 5A, which is published as sup-porting information on the PNAS web site) but not in parallelsamples from animals infected by RCAS-GFP (Fig. 5B).

Similar microscopic lesions can be found in mice carrying anMMTV-PyMT transgene (17, 18). Using a transgenic line (line634) known to manifest tumors relatively quickly, we comparedthe early lesions from RCAS-PyMT-infected and MMTV-PyMTtransgenic animals for proliferative and apoptotic indices and forexpression of genes associated with proliferation. Although theapoptotic index (measured by cleaved caspase 3) was very low inglands from both types of mice (data not shown and ref. 19), theproliferative indices were significantly higher in the lesions frominfected glands: 83% (�6%) of the cells in early lesions fromRCAS-PyMT-infected mice were positive for Ki-67, which labelscycling cells (Fig. 5E), but only 31% (�8%) were positive inhyperplastic lesions from MMTV-PyMT transgenic mice (Fig.5F). This difference does not appear to result from differentnumbers of cells with detectable levels of PyMT protein, becausePyMT was observed by immunohistochemical staining to bepresent in nearly all cells in early lesions induced by bothmethods (Fig. 5 C and D). Nor could this difference be attributedto the level of PyMT protein in individual cells in these earlylesions, because the PyMT signal strength, measured by immu-nohistochemical staining, appears to be weaker in early lesionsinduced by RCAS-PyMT (Fig. 5C) than in size-matched lesionsinduced by MMTV-PyMT (Fig. 5D).

Immunohistochemical staining was also used to examine theearly lesions for production of Cyclin D1 and c-Myc, two factorsknown to promote cell proliferation. Cyclin D1 was detected in67% (�10%) of cells in the early lesions induced by RCAS-PyMTviruses (Fig. 5G), but in only 34% (�12%) of cells in comparableearly lesions in MMTV-PyMT transgenic mice (Fig. 5H). Like-wise, c-Myc was detected in 44% (�8%) of cells in the formerlesions (Fig. 5I), but in very few cells in the latter (Fig. 5J). Wehave not determined whether the high fraction of cells contain-

Fig. 1. Generation and infection of MMTV-tva transgenic mice. (A) Diagramof the MMTV-tva transgenic construct. (B and C) TVA is produced in mammaryglands in MA mice. Immunohistochemical staining using rabbit antibodiesagainst TVA was used to detect TVA in mammary sections from pubertal MAmice (B) and nontransgenic littermate controls (C). (D and E) TVA is producedin mammary epithelial cells in MA mice. Mammary sections from pubertal MAmice were stained for TVA (green) and Keratin 8 (red) (D) or for TVA (red) and�-SMA (green) (E). (F and G) Mammary cells from pubertal MA mice aresusceptible to infection by RCAS. Primary mammary cells from MA mice wereinfected with RCAS-GFP viruses, stained for Keratin 8 (red) 3 days later, andphotomicrographed (F). Mock infection (G) was included as a negative con-trol. (H and I) Mammary cells in MA mice can be infected by intraductaldelivery of RCAS. RCAS-GFP viruses were injected via the nipple duct intopubertal MA mice (H) and littermate nontransgenic controls (I). Mammaryglands were collected 4 days later and stained for GFP by immunohistochem-istry. Original images were captured with a �40 objective.

Du et al. PNAS � November 14, 2006 � vol. 103 � no. 46 � 17397

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

July

20,

202

1

ing these proliferation-promoting proteins is a direct explanationfor the high proliferation index or an associated phenomenon.

Tumors Form Rapidly in MMTV-tva Mammary Glands Infected withRCAS-PyMT. The appearance of hyperplastic lesions with highproliferative capacity within a week after infection with RCAS-PyMT suggested that full-f ledged mammary cancers might de-velop rapidly in the infected mammary glands, despite therelatively small number of infected cells in each gland. Indeed,after infecting an additional 13 female mice from the MA line at6 weeks of age (injecting 107 units of RCAS-PyMT per gland viathe nipple duct into the left mammary glands numbers 2–4),multifocal tumors were detected with a surprisingly short medianlatency, 12.5 days, in all injected animals (Fig. 2A). As controls,we showed that MA females injected at the same age withRCAS-GFP viruses did not develop premalignant lesions ortumors within 3 months (Fig. 2 A and data not shown) and thattumors were not induced in non-tva-transgenic littermates in-

jected with the same dose of RCAS-PyMT virus (data notshown). We verified that the resulting mammary tumors werecomposed of cells infected with PyMT by Western blots, using arat antibody against PyMT to detect PyMT; a band with themobility expected for PyMT was detected in the lysates of tumorsfrom mice infected by RCAS-PyMT viruses, but not from lysatesof normal mammary glands (data not shown).

Mammary tumors appear in MMTV-PyMT transgenic mice ata median time of 5–25 weeks after birth, or 1–21 weeks afterpuberty onset, depending on the transgenic line (17, 19). TheKaplan–Meier plot of tumor-free survival for line 634, the mosttumor-prone of the MMTV-PyMT transgenic lines, is shown inFig. 2 A for comparison to the tumor-free survival plot for MAmice infected at 6 weeks of age with RCAS-PyMT. In thistransgenic line, the time from puberty to the appearance oftumors is similar to that observed from the time of infection ofMA mice with RCAS-PyMT; in other MMTV-PyMT transgeniclines, the time from puberty to tumor induction is longer.Expression of the MMTV-PyMT transgene, presumably begin-ning well before puberty, does not prevent expansion of theductal tree under the influence of the hormonal changes thataccompany puberty (data not shown) (18, 20), so that many moremammary cells in each of the 10 glands produce PyMT than inRCAS-PyMT-infected glands and hence are at risk of tumorinduction. It has been suggested that secondary oncogenicalterations at either the genetic or epigenetic level may benecessary for cells expressing the MMTV-PyMT transgene tobecome tumor cells (18, 20).

At this point, we cannot explain the difference in the kineticsin tumor appearance. PyMT expression levels might have somecorrelation with latency in existing MMTV-PyMT transgeniclines (17), but levels of PyMT protein produced by infection withRCAS-PyMT appear to be less than levels found in line 634, bothin early lesions [as determined by immunohistochemical staining(Fig. 5 C and D)] and in tumors [as determined by Westernblotting (data not shown)]. Thus, it appears that the efficientinduction of tumors by RCAS-PyMT virus reflects some strongoncogenic effect achieved by somatic delivery of PyMT, notsimply the concentration of PyMT protein. (We cannot excludea potential tumor-promoting role of the immune system thatmight be activated because of somatic introduction of PyMT, andwe cannot exclude the remote possibility that tumor formationin these MA mice was accelerated by a mutation of an endog-enous protooncogene or tumor suppressor gene that was inci-dentally induced when the tva transgene was integrated into thegenome, although spontaneous tumors were not observed inaged MA mice.)

Because tumors appear rapidly in infected mice, it seemsprobable that expression of PyMT is sufficient to transform atleast some of the few thousand RCAS-PyMT-infected cells ineach gland into tumor cells. We tested this possibility by judgingthe clonality of the tumors with respect to integration sites forRCAS-PyMT proviruses. Tumor DNAs were digested withHindIII (which leaves the PyMT insert in RCAS proviral DNAintact and generates one fragment detectable with a PyMT probefrom each integration site), and each tumor DNA samplegenerated multiple bands (Fig. 6, which is published as support-ing information on the PNAS web site). These findings are mostreadily explained by concluding that each RCAS-PyMT-inducedtumor arose from several independently infected cells, each witha single uniquely placed provirus. However, we cannot excludethe possibility that some of these tumors arose from a single cellwith multiple proviruses.

The tumors induced by RCAS-PyMT were relatively welldifferentiated, consisting of many acini and heterogeneous celltypes (Fig. 2B), resembling adenomyoepithelioma/carcinoma inhumans. These histological features are in sharp contrast to thepoorly differentiated tumors typically arising in mice that carry

RCAS-PyMT MMTV-PyMT

A

age (days)

tum

or-f

ree

surv

ival

(%

)pr

imar

y tu

mor

met

asta

sis

B

0

20

40

60

80

100

0 10 20 30 40 50 60 70

RCAS-PyMT

RCAS-GFP

MMTV-PyMT

Fig. 2. RCAS-PyMT induces rapid formation of mammary tumors in MA mice.(A) RCAS-PyMT or RCAS-GFP was injected via the nipple duct into left no. 2–4mammary glands (107 units per gland) in 13 and 10 6-week-old MA females,respectively. The infected mice were monitored for tumor appearance bypalpation. The time of infection is noted by an arrow on the Kaplan-Meiertumor-free survival curves. The survival curve of line 634 of MMTV-PyMTtransgenic mice (n � 10) is included for comparison, and the approximate timeof puberty onset is indicated by an arrowhead. (B) Representative photomi-crographs are shown for H&E-stained mammary tumors and lung metastasesinduced by RCAS-PyMT and from line 634 of MMTV-PyMT transgenic mice. A�20 objective was used.

17398 � www.pnas.org�cgi�doi�10.1073�pnas.0608607103 Du et al.

Dow

nloa

ded

by g

uest

on

July

20,

202

1

an MMTV-PyMT transgene (Fig. 2B and ref. 17). Furthermore,the cells in tumors induced by RCAS-PyMT appear larger thanthose in tumors from MMTV-PyMT transgenic mice (Fig. 2B).

The histopathology of RCAS-PyMT-induced tumors appearsremarkably similar to that of mammary tumors arising in trans-genic mice carrying the MMTV-Wnt-1 transgene (21, 22).MMTV-Wnt-1-induced tumors harbor heterogeneous cell types,including epithelial cells, myoepithelial cells, and cells producingputative progenitor cell markers such as Keratin 6 and Sca-1(22–25), and they have been proposed to arise from bipotentialprogenitor cells (22, 25–27). In contrast, MMTV-PyMT-inducedtumors lack myoepithelial cells and appear to arise from moredifferentiated cells that express the gene encoding whey acidicprotein (22, 24, 28) [although evidence of myoepithelial cellaccumulation has been noted in tumors arising in mice harboringa conditionally activated PyMT gene targeted into the �-actinlocus (29)]. Using immunohistochemical staining, Keratin 8 and�-SMA were detected in patchy patterns, suggesting that epi-thelial cells and myoepithelial cells were indeed present intumors arising in RCAS-PyMT-infected mice (Fig. 7 A–D, whichis published as supporting information on the PNAS web site).The myoepithelial nature of these �-SMA� cells was furtherconfirmed by staining for Keratin 14 (data not shown). More-over, a small subset of tumors in RCAS-PyMT-induced tumorsalso produced Keratin 6 and Sca-1 (Fig. 7 E and G). Collectively,these data indicate that mammary tumors arising in RCAS-PyMT-infected mice are heterogeneous, suggesting an origin inimmature or progenitor cells as in the MMTV-Wnt-1 transgenicmodel. Consistent with previous reports (22, 24), whereas Ker-atin 8 was found in nearly all cells in tumors arising in MMTV-PyMT-transgenic mice, �-SMA, Keratin 6, or Sca-1 was found invery few tumor cells in these transgene-induced tumors (Fig. 7D, F, and H).

To test whether RCAS-PyMT-induced tumors were able tometastasize, we generated 11 more tumor-bearing mice. Five of24 tumor-bearing mice had developed single pulmonary metas-tases when the largest tumor had reached 1.5 cm in diameter(Fig. 2B). This rate is lower than what is typically seen in micetransgenic for MMTV-PyMT (17). To verify that the primarymammary tumors could be transplanted, we dissociated tumorcells from RCAS-PyMT-infected mice and inoculated 106 tumorcells into cleared fat pads of syngenic (FVB) recipient mice.Transplants of all five tumors resulted in new tumors (data notshown).

Infection with RCAS Vectors Can Identify Genes with the Potential toAccelerate Tumorigenesis in Established Transgenic Models. Havingshown that delivery of an oncogene with RCAS vectors is aneffective means for testing its transforming potential, we askedwhether infection of mammary glands in an established trans-genic model for breast cancer with RCAS vectors could be usedto test genes for their ability to accelerate tumorigenesis. To thatend, we attempted to hasten the appearance of tumors inMMTV-Wnt-1 transgenic mice by infecting the mammary glandswith RCAS vectors carrying known oncogenes.

First, we crossed MMTV-Wnt-1 mice with MMTV-tva trans-genics, to create bitransgenic animals susceptible to infectionwith RCAS(A); using RCAS-GFP, we found that we could infect�0.04% (�0.03%) of mammary cells in the glands of 12-week-old bitransgenic animals (Fig. 3A). Presumably, this inefficientinfection reflects the distortion of the hyperplastic mammaryducts in MMTV-Wnt-1 transgenic mice (21). Despite this lowefficiency of viral infection, all seven 12-week-old bitransgenicmice infected with RCAS carrying a mutant cDNA version of theNeu (ErbB2�HER2) protooncogene (RCAS-Neu; see Materialsand Methods), a cellular gene frequently implicated in humanbreast cancers, developed palpable tumors within 3 weeks after

infection, much sooner than the reported tumor latency inMMTV-Wnt-1 transgenic animals (21).

This accelerating effect was caused by oncogene collaborationand was not simply an effect of the Neu oncogene, becauseinfection of MMTV-tva monotransgenic animals with RCAS-Neu produced tumors relatively inefficiently. A median time of6 months was necessary for this virus to induce mammary tumorsin 14 12-week-old MA mice (Fig. 3B) despite the fact that thesemice were stimulated (at the time of infection) with estrogen andprogesterone to make available more proliferative cells for viralinfection (Fig. 3A). This latency is a few months longer than whathas been reported for most transgenic mouse lines that expressan activated Neu gene (30–32).

Eight weeks after infection, necropsy was performed on theinfected bitransgenic animals. One hundred gross and occulttumors were detected in 26 injected glands, or 3.8 tumors perinfected gland (Fig. 3C). In contrast, only 2 tumors were foundin 28 noninjected glands, or 0.1 tumor per gland (Fig. 3C). TheHA epitope in the Neu protein was detected in 91 of 100 tumorsin infected glands, at an average of 3.5 RCAS-Neu� tumors perinfected gland; whereas only 9% tumors in these infected glandswere negative for HA (Fig. 3D). These results confirm collab-oration between Neu and Wnt-1 in mammary tumorigenesis (33)and demonstrate that this approach is a valuable tool forscreening for collaborative genetic alterations, while eliminating

A

C

0

2

4

6

8

10

injected uninjected

3.8

0.1

0

5

10

Neu+ Neu-

3.5

0.3

B

tum

or N

o./g

land

tum

or N

o./g

land

D

RCAS-Neu

0.0

0.2

0.4

0.6

0.8

1.0

MA/Wnt MA

0.52

0.04

GF

P(%

)

0

20

40

60

80

100

0 20 40 60age (weeks)

tum

or fr

ee (

%)

Fig. 3. Use of the RCAS-mediated gene transfer method for identifyingcollaborating oncogenes. (A) Influence of the MMTV-Wnt-1 transgene oninfection efficiency. Five 12-week-old mice bitransgenic for MMTV-tva andMMTV-Wnt-1 (MA�Wnt) were injected with RCAS-GFP viruses (107 units pergland). Four days after the infection, the fraction of GFP� cells was quantifiedby flow cytometry. For comparison, five MA mice were also infected withRCAS-GFP; these mice were injected s.c. with estrogen (1 �g per mouse) andprogesterone (1 mg per mouse) for 5 consecutive days starting 1 day beforethe viral infection to simulate the hyperplasia observed in MA�Wnt mice. (B)Tumor-free survival curve in MA mice infected by RCAS-Neu. Fourteen 12-week-old MA mice were stimulated with hormones as above, and six glands ineach mouse were infected by RCAS-Neu (106 units per gland). Their tumor-freesurvival was monitored by palpation. (C) Tumor induction in mice bitransgenicfor MMTV-tva and MMTV-Wnt-1. Twenty-six glands in seven 12-week-oldmice were injected with RCAS-Neu (106 units per gland). All infected mice werekilled 8 weeks later, and their mammary glands were examined grossly andmicroscopically to quantify the number of gross and occult tumors per in-fected or noninfected gland. (D) Quantitation of Neu� and Neu� tumors ininfected glands. One section of each infected gland collected above wasstained for HA to quantify the number of tumors expressing Neu.

Du et al. PNAS � November 14, 2006 � vol. 103 � no. 46 � 17399

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

July

20,

202

1

the time and cost required to create individual transgenic linesfor each gene.

Myoepithelial cells are rare in tumors arising in MMTV-Neutransgenic mice (22, 24). They were also uncommon in tumorsinduced in MA mice by RCAS-Neu (Fig. 8L, which is publishedas supporting information on the PNAS web site). Tumorsarising in mice bitransgenic for MMTV-Neu and MMTV-Wnt-1largely resemble the cellular phenotype of tumors arising in micemonotransgenic for MMTV-Wnt-1 (33). The Neu-expressingtumors (Fig. 8 A–C) in infected mice bitransgenic for MMTV-tvaand MMTV-Wnt-1 were comprised predominantly of Keratin8-positive epithelial tumor cells (Fig. 8B), although a muchsmaller subset of �-SMA-positive cells were still detected (Fig.8C). We do not yet know whether the decline in the number ofmyoepithelial cells suggests that the tumors originated from acell that was more differentiated than the source of tumors inmice monotransgenic for MMTV-Wnt-1 or bitransgenic for bothMMTV-Wnt-1 and MMTV-Neu. As expected, Neu-negativetumors arising in infected glands in these bitransgenic mice (Fig.8 D–F) had many myoepithelial tumor cells (Fig. 8F) in additionto epithelial cells (Fig. 8E), similar to those arising in mice thatwere not injected with RCAS-Neu (Fig. 8 G–I).

DiscussionWe have successfully adapted the TVA-mediated retroviral genedelivery method for introducing oncogenes into the mammarygland. This method overcomes the need to create individualtransgenic lines for in vivo expression of oncogenes and may bea valuable alternative to transgenic approaches for testing thetumorigenic potential of candidate genes, even though someoncogenes (Neu as opposed to PyMT) introduced into themammary gland by this method may induce tumors more slowlythan those expressed in conventional transgenic mice. Moreover,by using RCAS to introduce test genes into mammary glands thatare cancer-predisposed because of preexisting oncogenic trans-genes, genetic collaboration can be measured without the needto create and breed additional transgenic lines. Because theRCAS vectors introduce oncogenes into a small subset ofmammary cells in developmentally normal mammary glands, theresulting models may reveal important aspects of human breastcancer initiation that are otherwise difficult to explore, includinginteraction of oncogene-expressing cells with neighboring nor-mal epithelial cells and myoepithelial cells, remodeling of ex-tracellular matrix, and localized attraction of circulating cellsthat may assist in carcinogenesis.

RCAS-PyMT induced a stronger proliferative response andmore rapid tumorigenesis in the mammary gland than did theMMTV-PyMT transgene. The resulting tumors were also moreheterogeneous in cellular composition than tumors arising inMMTV-PyMT transgenic mice. Several reasons may account forthese differences.

First, the cellular environment is different: RCAS-PyMT-induced tumors arose in developmentally normal mammaryglands, whereas MMTV-PyMT-induced tumors evolved in a fieldof neighboring mutant cells in developmentally abnormal glands(17, 18, 20). Cells in mammary glands that have already devel-oped normally might be induced to proliferate more easily thancells in mammary glands that have not properly developed,because of alterations in cell differentiation and availability ofcirculating estrogen and other hormones.

Second, prepubertal expression of PyMT in MMTV-PyMTtransgenic mice may allow the dormant mammary cells tobecome partially adapted to that oncogene, whereas estrogensignaling in pubertal and older mice may help incoming RCAS-PyMT rapidly induce tumors. [Preliminary studies suggest thatestrogen deprivation indeed delays tumor development in MAmice infected by RCAS-PyMT (Z.D. and Y.L., unpublishedobservations).] Developmental influence on tumorigenic effects

of an oncogene has been reported: activation of ErbB2 duringembryogenesis has a much weaker transforming effect in mam-mary glands than activation of the same gene at a later devel-opmental time (34). A possible means to test whether thedevelopmental state of mammary glands has a significant influ-ence on tumor latency is to induce PyMT at different develop-mental times of mammary development in doxycycline-induciblePyMT transgenic mice.

Third, tumors in MMTV-PyMT transgenic mice have beensuggested to arise from more differentiated cells (22, 28), buttumors initiated by RCAS-PyMT may have an origin in progen-itor cells. This potential difference in the cellular origin oftumors may not be surprising, because RCAS-PyMT viruses maypreferentially infect progenitor cells, the cells that proliferatemost rapidly during puberty and are consequently most suscep-tible to successful infection. The cell cycle and growth machineryin progenitor cells may also be in a more activated state or maybe more easily activated upon stimulation by PyMT. Preliminarystudies suggest that RCAS viruses selectively infect mammarycells that are positive for both CD24 and CD49f (Z.D. and Y.L.,unpublished observations), which have been reported to beenriched for progenitor cells (35). Further supporting the pos-sibility of a progenitor origin of RCAS-PyMT-induced tumors isthe close similarity in histology and cellular composition be-tween RCAS-PyMT-induced tumors and MMTV-Wnt-1-inducedtumors, the latter of which seem to arise from progenitor cells(reviewed in ref. 26).

In conclusion, we have developed a method for testing thetumorigenic potential of candidate mammary oncogenes in vivo.Mouse models developed with this approach appear to havecharacteristics that differ from conventional models, may moreclosely recapitulate human breast cancer evolution, and may bevaluable for preclinical testing of new chemotherapeutic agents.

Materials and MethodsTransgenic Mice and Animal Care. The MMTV-tva transgenicvector was constructed in pGfa2-tva (6) by substituting for theGFAP promoter with a 1.2-kb MMTV LTR fragment releasedfrom pMMTV-p53 (36) by XbaI and XmaI. The resultingtransgene construct thus contains the MMTV LTR as thepromoter, the quail tva cDNA (850 bp) encoding a glycosylphos-phatidylinositol-linked form of TVA, and the mouse prota-mine-1 poly(A) signal. This 2.7-kb transgene fragment wasreleased from the transgenic vector by digestion with BglII andAatII and injected into pronuclei from FVB�N mice. Potentialfounders were screened by PCR using primers specific to tva andconfirmed by Southern hybridization using probes made fromthe transgene. MMTV-PyMT and MMTV-Wnt-1 transgenic micehave been described (17, 21). All animals used in this study wereon the FVB�N genetic background.

Virus Preparation and Delivery to the Mammary Gland. RCAS-PyMThas been described (37). RCAS-Neu contains a HA-tagged ratNeu cDNA insert with a Val-to-Glu point mutation of codon 664and truncations at both extracellular and intracellular domains(38). RCAS-GFP was a gift of Connie Cepko (Harvard MedicalSchool, Boston, MA). To produce RCAS viruses, DF-1 chickenfibroblasts (39, 40) were transfected with RCAS vectors by usingthe Superinfect transfection reagent (Qiagen, Valencia, Ca), andmaintained in DMEM supplemented with 10% FBS in humid-ified 37°C incubators supplemented with 5% CO2. Viruses in theculture supernatant were concentrated 100-fold by centrifuga-tion at 125,000 � g for 90 min, resuspended in DMEM contain-ing 10% FBS, and frozen in aliquots for titer determination andinfection of cells and animals. Virus titers were determined bylimiting dilution on DF1 cells. To infect mammary glands, femalemice were anesthetized and injected through intraductal injec-

17400 � www.pnas.org�cgi�doi�10.1073�pnas.0608607103 Du et al.

Dow

nloa

ded

by g

uest

on

July

20,

202

1

tion (15) with concentrated RCAS viruses in a 10-�l volume inconjunction with a tracking dye (0.1% bromophenol blue).

Tumor Harvest and Analyses. Mammary glands and tumors wereremoved and either fixed in 10% buffered formalin overnight at4°C or snap-frozen in liquid nitrogen. Fixed tissues were paraf-fin-embedded, and 3-�m sections were generated and placed onslides for H&E staining and�or immunostaining. Immunohisto-chemistry, immunofluorescence staining, and Western blottingwere performed as described (22, 41). Primary antibodies usedin these experiments include rabbit IgGs against GFP (Molec-ular Probes, Carlsbad, CA), TVA (a gift of Andy Leavitt,University of California, San Francisco) (42), Ki67 (NovoCastra,Newcastle, UK), Cyclin D1 (Lab Vision Corporation, Fremont,CA), c-Myc (sc-764; Santa Cruz Biotechnology, Santa Cruz,CA), Keratin 6 (all isoforms including a, b, and H-F; Covance,Richmond, CA) (43, 44), and PyMT (Ab-4; Calbiochem, SanDiego, CA). Also used were mouse monoclonal antibodiesagainst �-SMA (DAKO, Glostrup, Denmark) and GFP (BDBioscience, San Jose, CA); rat IgG against Sca-1 (15-5981-81;BD PharMingen, Franklin Lakes, NJ); a partially purified ratantibody against keratin 8 (ref. 45; Developmental StudiesHybridoma Bank, Iowa City, IA); and rat anti-PyMT ascitescollected from Brown Norwegian (B�N) rats that had been

injected i.p. with cultured B�N rat fibroblasts transformed withpolyoma virus (46). Southern hybridization was performed asdescribed (22).

We thank Drs. Jeffrey Rosen, Eric Holland, Daniel Medina, GaryChamness, and Adrian Lee for stimulating discussions and�or criticalreview of this manuscript; the Pathology Core Facility at the BreastCenter for tissue processing; the Transgenic Mouse Facility at BaylorCollege of Medicine for animal husbandry; Amy Chen and Lisa Garrattof the Transgenic Core Facility at National Human Genome ResearchInstitute, National Institutes of Health (Bethesda, MD) for assistance increating the transgenic mice; Dr. Dorothy Lewis and Jeff Scott of theBaylor College of Medicine FACS facility for FACS analysis; Dr. AndyLeavett for anti-TVA antibodies; Dr. Connie Cepko for the RCAS-GFPvector; and the laboratory of Dr. Margaret Neville (University ofColorado Health Sciences Center, Denver, CO) for demonstrating theintraductal injection technique to Y.L. This work was started at theNational Cancer Institute in Bethesda, MD and was supported in part byNational Institutes of Health Grant R01 CA113869 (to Y.L.), Project 5of Mouse Models of Human Cancers Consortium Grant U01CA084243-07 (to Y.L.; principal investigator: Dr. Jeffrey Rosen, BaylorCollege of Medicine), National Institutes of Health Grant P01CA94060-02 (to H.E.V.), National Cancer Institute Grant P50CA058183 (to Y.L.; principal investigator: Dr. C. Kent Osborne, BaylorCollege of Medicine), and U.S. Army Medical Research and MaterielCommand Grant BC030755 (to Y.L.).

1. Green JE, Hudson T (2005) Nat Rev Cancer 5:184–198.2. Li Y, Hively WP, Varmus HE (2000) Oncogene 19:1002–1009.3. Fisher GH, Orsulic S, Holland E, Hively WP, Li Y, Lewis BC, Williams BO,

Varmus HE (1999) Oncogene 18:5253–5260.4. Orsulic S (2002) Mamm Genome 13:543–547.5. Federspiel MJ, Bates P, Young JA, Varmus HE, Hughes SH (1994) Proc Natl

Acad Sci USA 91:11241–11245.6. Holland EC, Varmus HE (1998) Proc Natl Acad Sci USA 95:1218–1223.7. Holland EC, Hively WP, DePinho RA, Varmus HE (1998) Genes Dev

12:3675–3685.8. Orsulic S, Li Y, Soslow RA, Vitale-Cross LA, Gutkind JS, Varmus HE (2002)

Cancer Cell 1:53–62.9. Montaner S, Sodhi A, Molinolo A, Bugge TH, Sawai ET, He Y, Li Y, Ray PE,

Gutkind JS (2003) Cancer Cell 3:23–36.10. Lewis BC, Klimstra DS, Varmus HE (2003) Genes Dev 17:3127–3138.11. Lewis BC, Klimstra DS, Socci ND, Xu S, Koutcher JA, Varmus HE (2005) Mol

Cell Biol 25:1228–1237.12. Pao W, Klimstra DS, Fisher GH, Varmus HE (2003) Proc Natl Acad Sci USA

100:8764–8769.13. Fu SL, Huang YJ, Liang FP, Huang YF, Chuang CF, Wang SW, Yao JW (2005)

Biochem Biophys Res Commun 338:830–838.14. Rollini P, Billotte J, Kolb E, Diggelmann H (1992) J Virol 66:4580–4586.15. Nguyen D-A, Beeman N, Lewis M, Schaack J, Neville MC (2000) in Methods

in Mammary Gland Biology and Breast Cancer Research, eds Ip MM, Asch BB(Kluwer, New York), pp 259–270.

16. Ichaso N, Dilworth SM (2001) Oncogene 20:7908–7916.17. Guy CT, Cardiff RD, Muller WJ (1992) Mol Cell Biol 12:954–961.18. Lin EY, Jones JG, Li P, Zhu L, Whitney KD, Muller WJ, Pollard JW (2003)

Am J Pathol 163:2113–2126.19. Webster MA, Hutchinson JN, Rauh MJ, Muthuswamy SK, Anton M, Tortorice

CG, Cardiff RD, Graham FL, Hassell JA, Muller WJ (1998) Mol Cell Biol18:2344–2359.

20. Maglione JE, Moghanaki D, Young LJ, Manner CK, Ellies LG, Joseph SO,Nicholson B, Cardiff RD, MacLeod CL (2001) Cancer Res 61:8298–8305.

21. Tsukamoto AS, Grosschedl R, Guzman RC, Parslow T, Varmus HE (1988) Cell55:619–625.

22. Li Y, Welm B, Podsypanina K, Huang S, Chamorro M, Zhang X, Rowlands T,Egeblad M, Cowin P, Werb Z, et al. (2003) Proc Natl Acad Sci USA100:15853–15858.

23. Cui XS, Donehower LA (2000) Oncogene 19:5988–5996.

24. Rosner A, Miyoshi K, Landesman-Bollag E, Xu X, Seldin DC, Moser AR,MacLeod CL, Shyamala G, Gillgrass AE, Cardiff RD (2002) Am J Pathol161:1087–1097.

25. Liu BY, McDermott SP, Khwaja SS, Alexander CM (2004) Proc Natl Acad SciUSA 101:4158–4163.

26. Li Y, Rosen JM (2005) J Mammary Gland Biol Neoplasia 10:17–24.27. Shackleton M, Vaillant F, Simpson KJ, Stingl J, Smyth GK, Asselin-Labat ML,

Wu L, Lindeman GJ, Visvader JE (2006) Nature 439:84–88.28. Henry MD, Triplett AA, Oh KB, Smith GH, Wagner KU (2004) Oncogene

23:6980–6985.29. Politi K, Kljuic A, Szabolcs M, Fisher P, Ludwig T, Efstratiadis A (2004)

Oncogene 23:1558–1565.30. Guy CT, Cardiff RD, Muller WJ (1996) J Biol Chem 271:7673–7678.31. Muller WJ, Sinn E, Pattengale PK, Wallace R, Leder P (1988) Cell 54:105–115.32. Bouchard L, Lamarre L, Tremblay PJ, Jolicoeur P (1989) Cell 57:931–936.33. Podsypanina K, Li Y, Varmus H (2004) BMC Med 2:24.34. Andrechek ER, Hardy WR, Laing MA, Muller WJ (2004) Proc Natl Acad Sci

USA 101:4984–4989.35. Stingl J, Eirew P, Ricketson I, Shackleton M, Vaillant F, Choi D, Li HI, Eaves

CJ (2006) Nature 439:993–997.36. Godley LA, Kopp JB, Eckhaus M, Paglino JJ, Owens J, Varmus HE (1996)

Genes Dev 10:836–850.37. Holland EC, Li Y, Celestino J, Dai C, Schaefer L, Sawaya RA, Fuller GN

(2000) Am J Pathol 157:1031–1037.38. Bargmann CI, Weinberg RA (1988) EMBO J 7:2043–2052.39. Himly M, Foster DN, Bottoli I, Iacovoni JS, Vogt PK (1998) Virology

248:295–304.40. Schaefer-Klein J, Givol I, Barsov EV, Whitcomb JM, VanBrocklin M, Foster

DN, Federspiel MJ, Hughes SH (1998) Virology 248:305–311.41. Li Y, Podsypanina K, Liu X, Crane A, Tan LK, Parsons R, Varmus HE (2001)

BioMedCentral Mol Biol 2:2.42. Murphy GJ, Leavitt AD (1999) Proc Natl Acad Sci USA 96:3065–3070.43. Roop DR, Cheng CK, Titterington L, Meyers CA, Stanley JR, Steinert PM,

Yuspa SH (1984) J Biol Chem 259:8037–8040.44. Grimm SL, Bu W, Longley MA, Roop DR, Li Y, Rosen JM (2006) Breast

Cancer Res 8:R29.45. Kemler R, Brulet P, Schnebelen MT, Gaillard J, Jacob F (1981) J Embryol Exp

Morphol 64:45–60.46. Chen L, Wang X, Fluck MM (2006) J Virol 80:7295–7307.

Du et al. PNAS � November 14, 2006 � vol. 103 � no. 46 � 17401

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

July

20,

202

1