Somatic Mutation of p53 Leads to Estrogen Receptor α-Positive and -Negative Mouse Mammary Tumors with High Frequency of Metastasis

Approximately 70% of human breast cancers are estrogen receptor α (ERα)-positive, but the origins of ERα-positive and -negative tumors remain unclear. Hormonal regulation of mammary gland development in mice is similar to that in humans; however, most mouse models produce only ERα-negative tumors. In addition, these mouse tumors metastasize at a low rate relative to human breast tumors. We report here that somatic mutations of p53 in mouse mammary epithelial cells using the Cre/loxP system leads to ERα-positive and -negative tumors. p53 inactivation under a constitutive active WAPCrec in prepubertal/pubertal mice, but not under MMTVCre in adult mice, leads to the development of ERα-positive tumors, suggesting that target cells or developmental stages can determine ERα status in mammary tumors. Importantly, these tumors have a high rate of metastasis. An inverse relationship between the number of targeted cells and median tumor latency was also observed. Median tumor latency reaches a plateau when targeted cell numbers exceed 20%, implying the existence of saturation kinetics for breast carcinogenesis. Genetic alterations commonly observed in human breast cancer including c-myc amplification and Her2/Neu/erbB2 activation were seen in these mouse tumors. Thus, this tumor system reproduces many important features of human breast cancer and provides tools for the study of the origins of ERα-positive and -negative breast tumors in mice.


INTRODUCTION
Breast carcinogenesis requires multiple genetic changes including inactivation of tumor suppressor genes and activation of oncogenes (1). Mutations in p53 are observed in close to half of human cancers including breast carcinomas (2). p53 is also mutated in families with Li-Fraumeni syndrome in which early-onset female breast cancer is the most prevalent type of tumor (3). p53 is a transcription factor that regulates genes critical for cell cycle arrest and for apoptosis after genotoxic stress, thus preventing genome instability (4 -6). In addition, amplification and/or overexpression of c-myc, Her2/Neu/erbB2, and cyclin D1 oncogenes are seen in a significant portion of breast cancers (2,7).
p53 knockout mice are cancer prone and develop early-onset lymphoma and sarcoma (8,9) but rarely mammary tumors (10) because of early mortality. To circumvent this problem, p53 null mammary epithelium is transplanted into the fat pad of wild-type recipients and leads to the formation of breast tumors (11). Although this offers a potential model to study breast tumors, the influence of the transplantation process on carcinogenesis is not clear. Also, the transplanted cells are p53 null , whereas in human tumors somatic mutations are acquired in a subset of cells during tumor progression. Previously, Jonkers et al. (12) reported that no mammary tumor formation was observed in p53 conditionalmutant mice carrying K14Cre transgene. Therefore, conditional inactiva-tion of p53 in mouse mammary epithelial cells is necessary to generate a mouse model mimicking human carcinogenesis.
In addition to genetic mutations, steroid hormones play a critical role in breast carcinogenesis (13). About 70% of human breast cancers are estrogen receptor ␣ (ER␣)-positive and estrogen-dependent (14), and ER␣ and progesterone receptor (PR) expression is an important indicator of potential responses to hormonal therapy (15). However, the factors that control ER␣ expression in tumor cells are unknown. Thus far, most established mouse models seldom produce ER␣-positive mammary tumors (16). In a C3(1)/SV40 T-antigen-transgenic model, ER␣ expression decreases during early mammary tumor progression (from low-to highgrade mammary intraepithelial neoplasia and becomes undetectable in invasive tumors (17). In Brca1-and Brca2-linked mammary tumors, the majority of tumors show no detectable ER␣ expression (18 -20).
We have generated a mouse breast tumor model by using Cre/loxp method to specifically inactivate p53 in mammary epithelial cells. This conditional inactivation of p53 leads to ER␣-positive and -negative mammary tumors with a high rate of metastasis. We found that p53 inactivation during specific developmental stages critically determines ER␣ expression in mammary tumors. This breast tumor system provides a close model of the human disease and will be useful for both mechanistic and therapeutic studies of ER␣-positive breast cancer.

MATERIALS AND METHODS
Targeting Vector Construction. A 10.75-kb clone covering exons 1-10 of p53 was isolated from a 129sv mouse genomic library. The 3.4-kb XhoI-HindIII fragment containing exons 2-9 was subcloned into pBR322. A replacement-type targeting vector was made by inserting the first loxP site into a BamHI site located in intron 6. A neo-cassette flanked by two loxP sites was inserted into the PvuII site in intron 4. The resulting construct was cleaved with XhoI and HindIII, blunt-ended and subcloned into p2TK (27). The finished targeting construct is designated p53neoloxp 3 2TK.
MMTVCre Transgene Construction. The backbone of the MMTVCre transgene is pBSpKCR3 (28) containing part of the rabbit ␤-globin gene (the end of exon 2, intron 2, and exon 3), and the polyadenylation site. A 1.6-kb BglII-HindIII fragment composed of the Cre transgene with a nuclear localization signal was inserted within exon 3 of the globin gene of pBSpKCR3. A 1.5-kb HindIII-NheI fragment containing the MMTV-LTR was cut from pMAM (Clontech, Palo Alto, California) and subcloned 5Ј to exon 2 of the globin gene. The finished transgene construct is designated pMMTV ␤ Cre.
MMTVCre-Transgenic Mice Production. The 4.3-kb XhoI fragment containing the MMTVCre transgene was excised and purified. The transgenic founders were generated by microinjecting the MMTVCre transgene fragment into the male pronucleus of fertilized eggs derived from CB6F1 ϫ C57BL/6 intercrosses. Transgenic founders were identified by Southern analysis or PCR of tail DNA. The primers for the Cre transgene (363-bp amplified) were CreF (5Ј-GGTGTCCAATTTACTGACCGTACA-3Ј) and CreR (5Ј-CGGATCCGC-CGCATAACCAGTG-3Ј). All mice were maintained in accordance with the guidelines of Laboratory Animal Research of The University of Texas Health Science Center at San Antonio and Institutional Animal Care and Use Committee of University of California, Irvine.
Generation of Conditional p53 Mutant Mice. J1 embryonic stem cells were electroporated with SalI-linearized p53neoloxp 3 2TK and selected with G418 and 1-(2-deoxy-2-fluoro-␤-D-arabinofuranosyl)-5-iodouracil. Embryonic stem cells harboring homologous recombination were identified by Southern blotting using a 3Ј probe external to the targeting region. The neo-cassette was removed from targeted embryonic stem cells by transient expression of pSPCre. Of 231 clones analyzed by Southern blotting, two contained a recombination that removed only neo. These two clones were expanded and injected into C57BL/6 blastocysts. Chimeric males were mated with C57BL/6 females, and germline transmission of the mutant allele was verified by Southern and PCR analyses. Subsequently, p53 fp/fp MMTVCre mice were generated by crossing p53 fp/fp mice with MMTVCre mice. For PCR analysis, the following primers were used: primer x (5Ј-TGGGA-CAGCCAAGTCTGTTA-3Ј); y (5Ј-GCTGCAGGTCACCTGTAG-3Ј); z (5Ј-CATGCAGGAGCTATTACACA-3Ј); and p (5Ј-TACTCTCCTCCCCTCAATA-AGCTAT-3Ј). Primers "y" and "z" flank the loxP site in intron 6 and amplify a 119-bp fragment from wild-type p53 and 158 bp from the FP allele. Primer pair x/z amplifies a ϳ500-bp fragment from the deleted allele after Cre-mediated recombination. Primer pair p/q amplifies a 327-bp product in the wild-type allele as well as a 253-bp fragment for the pseudogene.
In vivo functional analysis of Cre recombinase in double-transgenic mice carrying Cre and R26R reporter transgenes. To evaluate Cre activity in vivo, Cre mice were crossed with the Rosa 26 reporter (R26R) strain (29). Mammary glands from double-transgenic Cre; R26R mice were collected at different developmental stages and stained with X-gal for lacZ expression (30).

Generation of the Floxed p53
Allele. We engineered the floxed p53 allele (designated p53 fp ) by inserting the loxP sites into introns 4 and 6 of p53 through a two-step process (Fig. 1, A and B). The phosphoglycerate kinase neomycin resistance cassette was subsequently removed by transient expression of Cre recombinase. Deletion of exons 5 and 6 led to an in-frame deletion of 99 codons (codons 123-221) that encode part of the DNA-binding domain. The presence of loxP sites in introns 4 and 6 did not interfere with the transcription of p53, and the full-length p53 protein was detected in p53 fp/fp mouse embryonic fibroblasts in the absence of Cre recombinase (Fig. 1C). A smaller protein product of expected mass of 39 kDa, designated p53 ⌬5,6 , was detected in p53 fp/fp mouse embryonic fibroblasts infected with Cre adenoviruses (Fig. 1C). IR-induced responses demonstrate that p53 ⌬5,6 is transcriptionally inactive and fails to increase p21 target gene transcription. In contrast to wild-type p53, mutant p53 ⌬5,6 protein was not stabilized after IR (Fig. 1D). No p53 protein was detected from mouse embryonic fibroblasts derived from p53-null (p53 Ϫ/Ϫ ) mice ( Fig. 1, C and D).
Characterization of Cre-Transgenic Mice. To introduce somatic p53 mutation in the mammary gland, different lines of mice expressing Cre under the control of MMTV-long terminal repeat, MMTVCre a , and MMTVCre b (collectively termed MMTVCre) were generated. In addition, to express Cre specifically in mammary luminal epithelial cells, WAPrtTACre-transgenic mice in which Cre is regulated by the whey acidic protein (WAP) promoter were also used (31). In the MMTVCre a line, Cre activity was restricted to the mammary gland, whereas in MMTVCre b and WAPrtTACre lines, Cremediated deletion was observed in many tissues ( Fig. 2A). In contrast to the founder mice, when the WAPrtTACre transgene was bred to floxed p53 or reporter mice (see below), its activity became independent of doxycycline and pregnancy, likely attributable to modification of the transgene, and/or the unstable nature of the multicopy transgene. The altered transgene will be referred to as WAPCre c hereafter.
Because transgene activity varies depending on promoter and insertion sites, Cre-transgenic mice were crossed with R26R mice expressing LacZ after the removal of the floxed stop sequence by Cre (29) to identify the cell types targeted by Cre-mediated recombination in the mammary gland ( Fig. 2B; Table 1; data not shown). In nulliparous MMTVCre a ; R26R mice, LacZ was detected in ϳ0.7% cells of mammary gland. The expression remained low during 1st pregnancy and reached ϳ2.9% at 2nd pregnancy in both luminal epithelial and myoepithlial cells but not in stromal fibroblasts and adipocytes. In MMTVCre b ; R26R mice, LacZ expression was also found in 5.6% of those cells in 2-week-old and ϳ20% in 7-week-old nulliparous mice. The expression increased to Ͼ60% in pregnant mice. The increased percentage of targeted cells in multiparous mice is expected, because MMTV promoter activities are up-regulated during pregnancy. In WAPCre c ; R26R mice, LacZ activities were robust with 66% positive cells in 1-week-old and Ͼ90% in 17-dayold and 6-week-old nulliparous and pregnant mice. Thus, there are significantly higher numbers of targeted cells in WAPCre c than in MMTVCre mice during prepubertal/pubertal stages.
Effects of p53 Inactivation on Spontaneous Mammary Carcinogenesis. As expected, these mice develop tumors with different spectrums. The most common tumor type is mammary tumor, and the majority of mammary tumor-bearing mice (60 -80%) had multiple primary mammary tumors ( Fig. 3; Table 2). Mammary tumor latency varied depending partly on numbers of targeted cells and parity of mice. Nulliparous p53 fp/fp MMTVCre a females developed mammary tumors between 14 and 24 months of age with a median tumor latency (MTL) of 17.5 months with almost complete penetrance (23 of 24 mice), whereas all multiparous p53 fp/fp MMTVCre a mice developed mammary tumors with a significantly shortened latency, between 11.5 and 24 months of age with a MTL of 15.5 months (P ϭ 0.004, generalized Wilcoxon statistic). Both larger numbers of targeted cells as well as pregnancy-mediated cell proliferation could contribute to the shortened MTL. Nulliparous p53 fp/fp MMTVCre b females developed mammary tumors between 6 and 14 months of age with a MTL of 10.5 months in nearly 50% of mice (9 of 19 mice), whereas multiparous p53 fp/fp MMTVCre b mice developed mammary tumors between 7 and 12 months of age with a MTL of 11 months in 60% of mice (12 of 20 mice). The MTL of p53 fp/fp MMTVCre b mice is significantly shorter than that of p53 fp/fp MMTVCre a mice (P Ͻ 0.001), indicating distinct mammary carcinogenesis kinetics in mice with 1 or 3% versus Ն20% of p53-mutated cells. However, there is no statistical difference in the median MTL between 20% targeted cells in nulliparous and 60% in multiparous p53 fp/fp MMTVCre b mice (P ϭ 0.29). The lower penetrance of mammary tumors in p53 fp/fp MMTVCre b was attributable to the presence of other tumor types leading to early death. In p53 fp/fp WAPCre c mice, mammary tumors developed between 8 and 12.5 months of age with a MTL of 9.5 months with high penetrance (13 of 14 mice). Consistent with the findings in p53 fp/fp MMTVCre b , increasing targeted cells to Ͼ90% did not shorten the MTL further (P ϭ 0.29). Of note, only one 23.5month-old multiparous heterozygous p53 ϩ/fp MMTVCre a mouse developed palpable mammary tumor in a cohort of mice between 20 and 26 months of ages (n ϭ 12; data not shown). The apparent long tumor  latency suggests that the internally truncated mutant p53 protein expressed under the endogenous promoter might not work in a dominantnegative manner.
Mammary tumors in all groups of mice are heterogeneous including adenocarcinoma, myoepithelial adenocarcinoma, adenosquamous carcinoma, and spindle cell tumor (Fig. 4, A-D). The majority of tumors were poorly differentiated invasive adenocarcinomas that share the most histopathological similarity with human tumors. In addition to mammary tumors, a few p53 fp/fp MMTVCre a mice also developed lymphoma caused by a very low level of Cre activity in the lymphocytes or a low spontaneous incidence in aged C57BL/6 ϫ 129sv mice. On the other hand, other tumor types were found in p53 fp/fp -MMTVCre b and p53 fp/fp WAPCre c mice caused by a broader Cre expression pattern (Table 2; Fig. 2A).
Importantly, up to half of p53 fp/fp MMTVCre a mice had mammary tumor metastasis either in lung and/or liver after gross examination ( Table 2; Fig. 4, E and F). Metastatic mammary tumor foci were also detected in p53 fp/fp MMTVCre b and p53 fp/fp WAPCre c mice ( Table 2).

Molecular Characterization of Mammary Tumors.
Estrogen is critical in the etiology of breast cancer and ER␣ mediates estrogen responsiveness in breast cancer. Significantly, 40% of the p53 fp/fp WAPCre c mice (n ϭ 15) had both ER␣-and PR-positive tumors (Fig. 4H), compared with 42 tumors from p53 fp/fp MMTVCre mice that were all ER␣-and PR-negative (Fig. 4G). The percentage of ER␣-positive cells in ER␣-positive tumors is over 90% whereas in ER␣-negative tumors, no ER␣-positive tumor cells were seen. The expression of ER␣ in tumors from p53 fp/fp WAPCre c but not p53 fp/fp -MMTVCre mice was further confirmed by Western blotting analysis (Fig. 4L). In contrast to ER␣, ER␤ was expressed in all tumors ( Fig.  4L and data not shown). The closely correlated expression of ER␣ and PR, a downstream target of ER, suggests that ER␣ is functional in ER␣-positive tumors. To further test this notion, adenoviruses carrying green fluorescent protein regulated by estrogen-response elements (EREs; 32) were used to infect tumor cells prepared from p53 fp/fp -MMTVCre and p53 fp/fp WAPCre c mice. After treatment with estradiol, green fluorescent protein-positive cells were detected in tumor cells from p53 fp/fp WAPCre c but not p53 fp/fp MMTVCre mice (Fig. 4M). Thus, ER␣ in the ER␣-positive p53 fp/fp WAPCre c tumors is transcriptionally active.
Deregulation of ER␣ expression during the premalignant stages of human breast carcinogenesis has been reported (33). Correspondingly, an increase of ER␣-positive cells as well as clusters of ER␣-positive cells (Fig. 4, J-K), in contrast to singularly distributed ER␣-positive cells in the normal gland, was observed in mammary intraepithelial neoplasia (34) but not in hyperplasia without atypia (Fig. 4I). Thus, similar to human breast cancer, there are multistep histopathological changes and alterations in the ER␣ expression pattern during the progression of mammary carcinogenesis in these models.
Frequent genetic changes and prognostic markers of human breast cancer have been identified. To test whether this mouse model parallels human breast cancer in these alterations, selected genes were    (Fig. 5). Amplification or overexpression of c-myc protooncogene is frequently observed in human breast cancer (35). In this model, c-myc amplification is found in ϳ35% of mammary tumors from p53 fp/fp MMTVCre mice (Fig. 5A). About two-thirds of tumors from p53 fp/fp MMTVCre and p53 fp/fp WAPCre c mice showed Neu/ erbB2 overexpression by immunostaining and Western blot analysis using anti-Neu/erbB2 and phosphotyrosine antibodies (Figs. 4H and 5B). Overexpressed Neu/erbB2 appears to be active because tyrosine residues of the receptors are phosphorylated (Fig. 5B). In addition, the activity of matrix metalloproteinase (MMP), matrix metalloproteinase 9 but not matrix metalloproteinase 2, was increased dramatically (Fig.  5C). Several prognostic markers [e.g., AYK1 (STK15), CDC25B, cyclin E2, and MCM6] identified recently through gene expression profiling in human breast cancer (36,37) also showed enhanced expression in these tumors (Fig. 5D).

DISCUSSION
Mice in a C57BL/6 ϫ 129sv background are normally resistant to mammary tumor development (38,39), but loss of p53 leads to a nearly complete incidence of mammary tumor with a high rate of metastasis to lung or liver. As the number of p53-inactivated cells increases, the MTL shortens. MTL reaches a plateau when Ն20% of the cells are targeted, indicating that there is a limit to the rate of breast carcinogenesis. Because ER␣-positive mammary tumors develop in prepubertal/pubertal but not in adult mice, the timing of p53 inactivation might be critical for determining ER␣ expression. Genetic changes (e.g., in c-myc and Her2/Neu/ erbB2) and prognostic markers [e.g., AYK1 (STK15), CDC25B, cyclin E2, and MCM6] associated with human breast cancer are also seen in these mouse tumors.
Despite the use of different promoters with Cre, both luminal epithelial and myoepithelial cells were targeted by Cre recombinase, as indicated by the Rosa26 reporter strain (29). Heterogeneous tumor types (adenocarcinomas, myoepithelial adenocarcinomas, adenosquamous carcinomas, and spindle cell tumors) were seen in all mouse strains (Fig. 4). The majority of tumors were poorly differentiated invasive adenocarcinomas, which were also the most similar to human tumors histopathologically. Nonmammary tumors were also found, probably attributable to Cre expression in other tissues (Table 2; Fig. 2A). About 50% of mammary tumors metastasize to lung or liver ( Table 2). Both histopathology and microarray analyses (data not shown) support that lung lesions are indeed tumor metastases. The metastasis frequency correlated with neither specific Cre-transgenic line nor tumor latency, consistent with previous observations (40). This tumor system recapitulates the high frequency of metastasis seen in advanced human breast cancer. By contrast, the experience of some investigators indicated that most mouse tumor models represent an early nonmetastatic stage of tumor development (41). Lung metastasis was also found in p53 Ϫ/Ϫ -BALB/c mammary gland transplant models (42); however, the precise metastasis rate is not clear. These mouse mammary tumors exhibit a pattern of mutation and gene dysregulation similar to human breast cancer. Amplification or overexpression of the c-myc and erbB2 proto-oncogenes is frequently observed in human breast cancer (7,35). About 35% of these mammary tumors were found to have c-myc amplification and two-thirds showed erbB2 overexpression. Matrix metalloproteinase 9 and cell cycle regulators such as AYK1 (STK15), CDC25B, cyclin E2, and MCM6 were up-regulated in these tumor cells. This is consistent with the recent results showing that overexpression of these genes in human breast cancer reflects poor prognosis (36,37). The high frequencies of c-myc amplification and overexpression of erbB2 and cell cycle regulators in p53-mutated tumors suggest that these genetic alterations have pivotal roles during tumor progression.
An inverse relationship between the number of targeted cells and MTL was observed. It is reasonable to assume that larger numbers of targeted cells are more likely to acquire critical genetic changes leading to tumor development with shorter latency. Indeed, similar genetic (e.g., c-myc) and gene expression changes (e.g., Neu/erbB2, AYK1, CDC25B, cyclin E2, and MCM6) were seen at high frequencies in tumors of both short and long latency. Intriguingly, when the number of targeted cells exceeded 20%, MTL did not shorten further. To explain this phenomena, we assume that cancer initiation depends in part on the accumulation of a certain number of key mutations (43). If the key rate-limiting process was simply the accumulation of one additional mutation, then after the average cell has gone through T rounds of cell division, the probability of obtaining the key mutation is NuT, where N is the number of target cells with the predisposing mutation and u is the mutation rate per cell division. If there are two rate-limiting mutational steps, then the probability of obtaining both mutations is, from the gamma distribution, approximately N(uT) 2 /2. If we suppose u is approximately 10 Ϫ6 , and there are about T ϭ 10 rounds of cell division, then saturation would happen when N is of the order 10 10 , which is probably much higher than the actual number of cells at saturation. Thus, to explain the saturation kinetics, there are two alternatives. First, the mutation rate per cell division may be higher than 10 Ϫ6 . This may occur because mutations of p53 increase the mutation rate per cell division. Second, some precancerous clonal expansion may be occurring, which would increase the number of rounds of cell division and therefore decrease the number N of initial predisposing target cells needed to achieve saturation. On the basis of results of the MMTVCre b line, it can be concluded that a saturation kinetic is reached by targeting ϳ20% of mouse mammary epithelial cells. In the WAPCre c line, it is likely that a different population of mammary epithelial cells is targeted; however, the tumor kinetics is similar to that of MMTVCre b . In addition to the number of targeted anti-erbB2 (Neu) and phosphotyrosine (p-Tyr) antibodies, truncated p53 protein (p53 ⌬5,6 ), and a loading control (p84). T, mammary tumors; N, normal mammary gland control; P, mammary tumor from an MMTV-Neu-transgenic mouse. C, changes in the relative activity of matrix metalloproteinase 9 (MMP9) and MMP2 in mammary tumors assayed by gelatin zymography. The value was normalized as a ratio to the activity of normal mammary gland. D, changes in the expression of several prognostic markers in mammary tumors. The relative levels of mRNAs of AYK1, CDC25B, cyclin E2, and MCM6 were analyzed by reverse transcription-PCR. Each value was normalized using the amount of ribosomal protein S16 reverse transcription-PCR product and was expressed as a ratio to the amount of mRNA in normal mammary gland. cells, it is plausible that the tumor latency might also be reflective of the type of cells targeted. Because multiple cell types are targeted in the models described here, the contribution of cell numbers and types of cells cannot be clearly differentiated.
On the basis of microarray profiling analysis, human breast cancers can be classified into five subtypes (44). It is not clear whether these heterogeneous tumor types are originated from different cells of the mammary gland or specific cancer-initiating cells are targeted and heterogeneity develops subsequently during tumor progression. Recent studies have revealed that a population of breast cancer cells possess stem cell-like properties (45). However, it remains to be studied whether cancer-initiating cells of ER␣-positive and -negative tumors are different. In WAPCre c mice both ER␣-positive and -negative mammary tumors were found, although mutations of p53 in MMTVCre mice resulted in only ER␣-negative tumors. Parity may not be relevant because ER␣-positive tumors were found in both nulliparous and parous p53 fp/fp WAPCre c mice. It is plausible that ER␣positive stem cells, in addition to ER␣-negative stem cells, are targeted in WAPCre c mice. In contrast, in MMTVCre mice, only ER␣negative stem cells are targeted. Because nearly 90% of cells are LacZ-positive during second pregnancy in MMTVCre mice (data not shown), this might suggest that only a small population of cells give rise to ER␣-positive tumors in these adult parous mice. The Cre transgenes are active at different developmental stages in WAPCre c and MMTVCre mice. Whether developmental stages affect the abundance of ER␣-positive progenitor cells is not clear. Because it is feasible to isolate ER␣-positive epithelial cells from normal mammary glands and tumors (32), 8 molecular mechanisms underlying ER␣-positive and -negative mammary carcinogenesis can be systematically addressed using this model.