|Home | About | Journals | Submit | Contact Us | Français|
We characterized the novel NRL-transforming growth factor alpha (NRL-TGFα) transgenic mouse model in which growth factor - steroid receptor interactions were explored. The NRL promoter directs transgene expression to mammary ductal and alveolar cells and is nonresponsive to estrogen manipulations in vitro and in vivo. NRL-TGFα mice acquire proliferative hyperplasias as well as cystic and solid tumors. Quantitative transcript analysis revealed a progressive decrease in estrogen receptor alpha (ER) and progesterone receptor (PR) mRNA levels with tumorigenesis. However, ER protein was evident in all lesion types and in surrounding stromal cells using immunohistochemistry. PR protein was identified in normal epithelial cells and in very few cells of small epithelial hyperplasias, but never in stromal or tumor cells. Prophylactic ovariectomy significantly delayed tumor development and decreased incidence. Finally, while heterozygous (+/−) p53 mice did not acquire mammary lesions, p53+/− mice carrying the NRL-TGFα transgene developed ER negative/PR negative undifferentiated carcinomas. These data demonstrate that unregulated TGFα expression in the mammary gland leads to oncogenesis that is dependent on ovarian steroids early in tumorigenesis. Resulting tumors resemble a clinical phenotype of ER+/PR−, and when combined with a heterozygous p53 genotype, ER−/PR−.
Ovarian hormones and growth factors influence mammary carcinogenesis. Primary breast tumors are categorized according to the presence estrogen receptor alpha (ER), and when expressed, selective estrogen receptor modulators such as tamoxifen are efficacious treatment options. Progesterone receptor (PR) expression also is an important clinical biomarker; ER+/PR− breast tumors are more aggressive and more likely to resist tamoxifen treatment compared to ER+/PR+ tumors which may be related to coincident elevated growth factor activity, often involving the EGFR family. Twenty-five percent of breast cancer cases are ER+/PR− (Arpino et al., 2005).
One EGF-like ligand, TGFα, frequently is over-expressed in primary breast cancers in women (Rudland et al., 1995). TGFα is a potent mitogen and, when secreted, stimulates synthesis of its receptor, ErbB1 (EGFR, HER1). TGFα binds to ErbB1, which subsequently dimerizes with one of the four ErbB receptors (ErbB1, ErbB2/HER2/neu, ErbB3 and ErbB4). ErbB1 and ErbB2 are often upregulated in refractory breast cancers and more often found expressed in ER− or ER+/PR− tumors (Holbro et al., 2003; Gee et al., 2005). These observations have led to breast cancer studies involving EGF ligand and receptors and ovarian hormone signaling.
Although there are several in vitro studies of crosstalk between ErbB1 ligands, ER and PR (Pierson-Mullany et al., 2003; Gururaj et al., 2006), in vivo models are sparse. While TGFα-induced mammary tumori-genesis has been described using the WAP, MMTV and MT promoters (Humphreys and Hennighausen, 2000; Rose-Hellekant and Sandgren, 2000), estrogen sensitivity and/or the expression pattern of these traditional promoters has limited their application. MMTV-transgene expression is uniform at 6 weeks, but becomes focally expressed by 15 weeks (Gunther et al., 2002), and progesterone activates this promoter (Otten et al., 1988). WAP promoter activity is upregulated during estrus (Robinson et al., 1995). MT directs only low mammary transgene expression (Sandgren et al., 1995). To study ovarian steroid interaction with TGFα during mammary tumorigenesis in transgenic mice, a promoter that would direct robust mammary transgene yet remain nonresponsive to ovarian hormones was required.
The NGAL homolog, neu-related lipocalin (NRL), is highly expressed in rat mammary glands and does not fluctuate with the estrous cycle (Stoesz and Gould, 1995), suggesting promoter independence of ovarian steroids. We therefore generated several lineages of NRL-TGFα transgenic mice. This promoter was not responsive to 17β-estradiol (E) in in vitro reporter assays. In addition, NRL-directed transgene expression was uniform in mammary epithelium in adulthood and not affected by E implant or ovariectomy. NRL-TGFα expression led to altered mammary development and ER+/PR− end stage neoplasms. ER but not PR protein was present in hyperplasias and tumors, mRNA levels progressively declined with tumorigenesis. The NRL-TGFα mouse is an in vivo model in which to evaluate the molecular and biological peculiarities of clinical ER+/PR− tumorigenesis and test prevention and treatment strategies of this resistant breast tumor type.
To study tissue distribution of NRL transgene expression, mice bearing the marker gene human placental alkaline phosphatase (NRL-hPAP) were generated. In two lineages that transmitted the transgene to offspring, hPAP staining was uniformly distributed in mammary epithelium (Figure 1a–e). Most other tissues were negative, although hPAP staining was found in epithelium of the reproductive tract, bile ducts, and in one line, pancreatic ducts.
To study the effects of NRL-TGFα tumorigenesis, we generated seven founder NRL-TGFα mice. Six founders transmitted transgene to offspring; four lineages were observed for mammary lesion development. As shown in Figure 2, virgin females acquired large mammary lesions with mean latencies of 9–21 months with an incidence of 100% in all but one line. Pregnancies reduced tumor latency, and all parous females developed lesions. Tumors were found rarely in other tissues and mice seldom died of causes unrelated to mammary tumors. Females were highly fertile and able to lactate. However, some males were infertile. Investigation of reproductive tracts revealed moderate hyperplasia in some secondary sex glands, but any relationship to fertility is unknown. Most mammary tumors developed in cranial glands, although hyperplasias were found in all glands preceding and concurrent with tumor development. End stage glands, collected when tumor diameter reached 1.5 cm, often consisted of a complex of heterogeneous histotypes (Table 1), including proliferative lobular and ductal hyperplasias (Figure 1g–i) and complex cystic (Figure 1j and k) and solid adenomatous tumors (Figure 1l and m).
To evaluate the role of hormones in transgenic models of mammary carcinogenesis, hormone insensitive promoters are required. Previously, we have found the 3 kb proximal NRL promoter fragment insensitive to prolactin (Rose-Hellekant et al., 2003). We assessed this fragment for E sensitivity in vitro using a luciferase reporter assay and found it not activated by E, in contrast to the ERE-containing oxytocin promoter (Figure 3a). However, the progesterone agonist, R5020, activated NRL approximately sixfold.
To confirm NRL insensitivity to E in vivo, we measured mammary NRL-TGFα transgene expression and compared it to WAP-TGFα expression. RNA was isolated from females in diestrus, d14 post-ovariectomy, d21 post- E implant or d18 pregnancy. Although WAP-TGFα transgene expression was down-regulated significantly with ovariectomy, NRL-TGFα expression was not affected by ovariectomy or E implant. Transgene expression in both models was upregulated strongly with pregnancy, consistent with the progesterone effect on NRL promoter activity.
Because breast cancers display an inverse relationship between growth factor and ER levels, we were interested in determining if chronic TGFα exposure altered mammary ER and PR expression. ER and PR mRNA levels appear to be reduced during tumorigenesis; tumor-containing glands had significantly lower levels of ER (sevenfold) and PR (40-fold) than normal glands (P<0.05), while hyperplastic glands were intermediate (Figure 4a and b). Higher TGFα transgene levels (26-fold) were found in tumor-containing glands than in transgenic glands without lesions coincident with increased epithelial content during tumorigenesis. However, normalization to epithelial composition was precluded by concomitant alterations in glandular cytokeratin content.
As mRNA levels may not reflect protein levels, we immunolocalized ER and PR in NRL-TGFα glands (Figure 1n–u). We found ER staining in both epithelial and stromal cells. Although ER mRNA levels were reduced in glands with hyperplasias and tumors, ER protein immunoscores were similar in ducts (81.9±41.5), lobules (90.4±34.9) and hyperplasias, (101.8±32.5, mean±sd., Figure 1o). In addition, most tumor cells displayed ER (Figure 1p and q). Between 0 and 23% of ducts and lobules contained PR, but PR was not found in nonepithelial cells (Figure 1r). Occasional PR+ cells were observed within small lobular hyperplasias (Figure 1s) but were undetectable in larger hyperplasias and tumor cells (Figure 1t and u), consistent with reduced PR mRNA levels found during tumorigenesis.
As estrogen has been implicated in breast carcino-genesis, we tested the effect of ovariectomy on TGFα tumorigenesis. Ovariectomy at 16 weeks dramatically increased tumor latency by ~50% (Figure 5). Three of nine ovariectomized mice were euthanized between 12.4–20.1 months prior to palpable mammary tumors and had the following abnormalities: liver or lung tumor or enlarged gall bladder.
The tumor suppressor gene, p53, maintains healthy tissues by suppressing proliferation of aberrant cells including those with carcinogenic potential. Mice with monoallelic p53 are prone to multiple cancers, usually succumbing to lymphoma without primary mammary disease (Jacks et al., 1994). We wished to determine if mammary TGFα expression would alter the p53+/− disease profile by promoting mammary cancer development.
Because p53+/− and NRL-TGFα mice are maintained in different inbred strains (C57BL/6 and FVB/N, respectively) we compared tumorigenesis in (FVB:B6)F1s. Six NRL-TGFα mice and 10 TGFα × p53+/− mice were monitored for approximately 2 years. Three of six TGFα(FVB:B6)F1 mice did not develop mammary tumors in spite of advanced age (23.7±0.7 months, Figure 6a), starkly different than tumor incidence in the FVB/N strain. In addition, in three mammary tumor bearing TGFα(FVB:B6)F1 mice latency was more than double that observed in the FVB/N strain. This influence of genetic background is similar to our report in WAP-TGFα mice (Rose-Hellekant et al., 2002). Despite a reduced tumor incidence and longer tumor latency in (FVB:B6)F1 versus FVB/N mice, NRL-TGFα induced tumors were histologically similar. However, when combined with p53+/− tumor incidence increased to 70% and latency was shortened significantly compared to TGFα (FVB:B6)F1 mice (P=0.0002). Unlike TGFα tumors, TGFα × p53+/− tumors were high grade and solid (Figure 6b) of differing histotypes:one lymphosarcoma, one squamous cell carcinoma, one fibro-sarcoma, three undifferentiated carcinomas, and one cystic papillary tumor similar to TGFα tumors. In stark contrast to TGFα tumorigenesis, dilated ducts or hyperplasias were not found at end stage in TGFα × p53+/− glands and tumors in TGFα × p53+/− were ER−/PR− (immunohistochemistry; data not shown). These data demonstrate that TGFα overcomes p53-mediated tumor suppression in cells with an inherently unstable chromosomal environment resulting from monoallelic p53.
Despite the importance of TGFα and ErbB-mediated pathways in refractory clinical breast cancers, in vivo models to explore these relationships during tumorigenesis have not been available. Here, we employed the novel NRL promoter, which drives uniform expression in mammary epithelial cells independently of estrogens and ovariectomy, to demonstrate that TGFα-induced tumorigenesis was strongly moderated by prophylactic ovariectomy. Although ER transcripts declined with tumorigenesis, end stage lesions remained ER+, but did not express PR. The declining but continued ER expression and associated lack of detectable PR observed in these lesions models the class of aggressive ER+/PR− breast tumors in women that prove refractory to hormonally-based therapeutics, and which often contain higher levels of growth factor ligands and/or receptors compared to ER + /PR + tumors (Arpino et al., 2005).
The NRL-TGFα model shares many features with ER+/PR− clinical breast cancers. ER is present in epithelial and nonepithelial mammary cells, increasing in hyperplasias (Lee et al., 2006). In addition, reduction in ER and, more radically, PR transcripts in NRL-TGFα mice parallels clinical findings in which breast cancers expressing TGFα and ErbB1 had reduced ER (Bieche et al., 2003).
Despite diminishing ER mRNA in NRL-TGFα mice, prophylactic ovariectomy significantly reduced tumor incidence and increased latency. Early dependence on ovarian hormones despite elevated growth factor expression contrasts with findings in MMTV-neu/ErbB2 mice; ovariectomy at similar ages did not alter tumor incidence or latency (Hewitt et al., 2002) suggesting constitutively active ErbB2 exerts more powerful growth/survival signals and early ovarian steroid independence, compared to TGFα mediated ErbB1 signaling. Cellular and tissue constraints which attentuate biological responses to TGFα may be responsible. Recent findings revealing distinctive molecular profiles for ER− tumors containing ErbB1 versus ErbB2 (Creighton et al., 2006) may shed light on these differences.
Cystic papillary tumors in NRL-TGFα mice resemble those of other TGFα transgenic mice (Humphreys and Hennighausen, 2000; Rose-Hellekant and Sandgren, 2000; Arendt et al., 2006). Concurrent expression of heterozygous p53 results in aggressive solid tumors of varying types reflecting the tumor spectrum of p53+/− BALB/c mice (Blackburn et al., 2003; Jerry et al., 2000; Kuperwasser et al., 2000), although relative proportions cannot be determined due to low numbers of p53+/− mice in the present study. The B6 background delayed tumor latency in NRL-TGFα mice, similar to that observed previously in WAP-TGFα mice (Rose-Hellekant et al., 2002) and p53+/− (BALB/c:B6) F1 mice (Jerry et al., 2000), suggesting that B6 alleles protect against mammary cancer. The increased incidence, shortened latency and increased tumor grade conferred by p53 heterozygosity in the NRL-TGFα model is similar to other transgenic mice expressing mammary oncogenes concurrently with heterozygous, null or mutated p53 (e.g. Wnt-1, c-myc, c-neu, ras, DES, BRCA1 and BRCA2; reviewed by Blackburn and Jerry, 2002). In combination with heterozygosity at the p53 locus, TGFα tumors were ER−, consistent with accelerated tumor progression facilitated by genomic instability or selective progression of a distinct tumor precursor subpopulation. Both ER+ and ER− tumors have been observed in other mouse models with one or more dysfunctional p53 alleles (Medina and Kittrell, 2003; Zhang et al., 2005).
ER and PR immunohistochemical status of NRL-TGFα tumors is similar to WAP-TGFα tumors (unpublished), and tumors from MMTV-TGFα × MMTV-c-myc bitransgenic mice also were ER+ using ligand binding assay (Amundadottir et al., 1995). The presence of ER in end stage NRL-TGFα tumors suggests that estrogen may continue to be an important mitogen even at later stages of tumorigenesis, although the sensitivity of tumor growth to estrogens or antiestrogens has not been tested in this model. However, the 40-fold reduction in PR mRNA and lack of detectable protein observed in these tumors strongly suggests that progesterone plays little or no role in established tumors, which may reflect tumor cellular origin or a cause-effect relationship between changes in ER status with tumor progression.
In summary, NRL-TGFα mice have the following characteristics: (1) Females develop proliferative hyperplasias, complex cystic and solid adenomatous tumors similar to clinical breast lesion histotypes. (2) Most early and end stage lesions are ER+, but become PR− very early, thus resembling the ER+/PR− tumor type found in 25% of women with breast cancer. (3) Tumor development can be impaired by prophylactic ovariectomy. (4) Finally, coupling a TGFα growth environment with monoallelic p53 results in aggressive ER− carcinomas. These data indicate not only a role for TGFα in breast carcinogenesis, but suggest an intricate relationship between TGFα, ER and PR in the NRL-TGFα mouse model reflecting many aspects of ER + /PR− breast carcinogenesis in women, and thus can be used to dissect underlying mechanisms of tumorigenesis in this resistant tumor type.
The 3.0 kb NRL promoter fragment was cloned upstream of 5.3 kb of hPAP coding sequence fused to the SV40 poly-adenylation signal (Kisseberth et al., 1999) or the rat TGFα minigene (Sandgren et al., 1990), and transgenic FVB/N mice generated as described (Brinster et al., 1985). NRL-TGFα and NRL-hPAP mice were identified, respectively, by PCR of tail DNA (Grippo and Sandgren, 2000) using primers spanning the NRL promoter and exon 1 of the rat TGFα gene (Table 2) or by enzyme histochemistry using the substrate 5-bromo-4-chloro-3-indolyl phosphate (BCIP) and confirmed by PCR (Kisseberth et al., 1999). Mice were housed in AAALAC accredited facilities with 14 h/d light, ad libitum Lab Diet 5015 and water. All procedures were IACUC approved. Certain NRL-TGFα lines are designated as: line 1372-1, TgN(Nrl-tgfa) 29EPS; line 1385-7 TgN(Nrl-tgfa)25EPS. Transgenic mice were evaluated as heterozygotes.
To evaluate the effect of genetic background, NRL-TGFα mice in the FVB/N strained were crossed once into the C57BL/6 strain generating (FVB:B6)F1 mice. To evaluate the effect of heterozygous p53 (p53+/−) on TGFα tumorigenesis, tumor phenotype in NRL-TGFα (FVB:B6)F1 mice was compared to (FVB:B6)F1 mice carrying both the TGFα transgene and p53+/− (Jacks et al., 1994).
Glands were fixed in 10% formalin for 18–24 h, or 4% paraformaldehyde at 4?C for 4 h, ethanol dehydrated, paraffin embedded, sectioned and stained with hematoxylin and eosin (H&E). hPAP expression was detected using BCIP histochemistry. For immunohistochemistry deparaffinized slides were subjected to 0.5% H2O2 in methanol, boiled in 0.1 m Tris, pH 9.0 or 0.1 m citrate buffer pH 6.0 for ER and PR, respectively, then blocked with 0.5% milk in phosphate buffered saline and incubated overnight with primary antibodies (1:1000 ER, MC-20, Santa Cruz Biotechnology Inc., Santa Cruz, CA, USA; 1/750 PR, #A0098, DakoCytomation, Carpinteria, CA, USA), rinsed and incubated with secondary antibody (BioGenex, San Ramon, CA, USA), rinsed and incubated with horseradish peroxidase streptavidin conjugate and developed with 3,3′ diaminobenzidine and counterstained with hematoxylin. Uteri were used as positive control tissues (not shown). Inactivated blocking peptide for ER (#sc-542P; Santa Cruz Biotechnology) and omitted primary antibody for PR were used as negative controls.
To evaluate NRL promoter responsiveness to E in vitro, a 3015 bp proximal promoter fragment (Stoesz and Gould, 1995; Rose-Hellekant et al., 2003) was cloned upstream of pGL3basic containing a luciferase reporter (NRL-Luc). E responsiveness also was evaluated in pGL3Basic and pOT-ERE (oxytocin promoter fragment containing an estrogen response element, ERE, upstream from the luciferase reporter; Adan et al., 1992). ERα (pSG5-ER, Green et al., 1988) was cotransfected into CHOK1 cells with luciferase and β-gal expression constructs, and responsiveness to 10 nm E determined (Gutzman et al., 2005). Responsiveness to the progesterone agonist, R5020 (10 nm, Perkin Elmer Life and Analytical Sciences, Inc., Waltham, MA, USA), was compared to pGL3basic containing 4 progesterone response element copies (PgRE; gift from Chinghai Kao) transfected into T47D cells using Lipofectamine2000 (Invitrogen, Carlsbad, CA, USA). To evaluate ovarian hormone responsiveness on mammary TGFα expression, 4–5.5 month old transgenic mice (N=4–5) were ovariectomized or implanted with E pellet (1.5 mg; Innovative Research of America, Sarasota, FL, USA) and glandular levels measured in diestrus or d18 pregnant mice. TGFα transcripts were normalized to endocytokeratin 18 (mcK18) to indicate epithelial content, using QPCR methods (Rose-Hellekant et al., 2002). Data were analysed using one-way analysis of variance (P≤0.05).
A SPOT camera, and MetaMorph Imaging and Adobe Photoshop 7.0 software were used to analyse epithelial morphology at 400 ×. Lesions were classified for histotype, ER (>10% ER+) or PR (>0% PR+), then assigned an average intensity score (1–3). Normal ductal epithelium was scored as 3. Immunoscores=intensity score × % immunopositive and ranged from 0 to 300.
Mammary RNA was isolated and 2.5 µg mRNA reverse-transcribed using oligo(dT) (StrataScript First-Strand Synthesis System, Stratagene). Reverse transcriptase-free and RNA-free controls were included. PCR was performed using 2 × Brilliant SYBR Green QPCR Master Mix (Stratagene), and the primers shown in Table 2. Average crossing threshold (Ct) values were calculated from triplicate cDNA samples. Reaction efficiencies were calculated from standard curves obtained from Ct data of serially diluted cDNA (Livak and Schmittgen, 2001). Outlier samples were identified using BestKeeper analysis (Pfaffl et al., 2004). Relative gene expressions were normalized to the housekeeping gene, PUM1, and calculated as Ct(PUM1)–Ct(gene)2 and significant differences determined using relative expression software tool (Pfaffl et al., 2002).
The authors greatly appreciate the assistance of Dr Donald Kundel, MD for histopathology evaluations, and Dr Lisa Arendt for critical manuscript reading. We also thank Chinghai Kao for graciously providing the PgRE construct. This work was supported by NIH grant K01-RR00145, University of Minnesota Medical Foundation (TARH), R01-CA64843 (EPS), R01-CA78312 (LAS), R01-CA58328 (MNG) and ES09090.