|Home | About | Journals | Submit | Contact Us | Français|
Mammary epithelia are composed of luminal and myoepithelial/basal cells whose neoplastic transformations lead to distinct types of breast cancers with diverse clinical features. We report that mice deficient for the CDK4/6 inhibitor p18Ink4c spontaneously develop ER-positive luminal tumors at a high penetrance. Ink4c deletion stimulates luminal progenitor cell proliferation at pubertal age and maintains an expanded luminal progenitor cell population throughout life. We demonstrate that GATA3 binds to and represses INK4C transcription. In human breast cancers, low INK4C and high GATA3 expressions are simultaneously observed in luminal A type tumors and predict a favorable patient outcome. Hence, p18INK4C is a downstream target of GATA3, constrains luminal progenitor cell expansion and suppresses luminal tumorigenesis in the mammary gland.
Breast cancer is heterogenous with tumors pathologically distinct and diverse in their responsiveness to treatment. We show that the CDK inhibitor p18INK4c gene is repressed by GATA3, a transcription factor specifying mammary luminal cell fate, and that low INK4C and high GATA3 expressions are associated with human luminal A type tumors and with better patient survival. p18Ink4c null mice have an expanded luminal progenitors at a young age and throughout life and develop ER+ luminal tumors at high penetrance. These results identify the escape of luminal progenitors from quiescence as a rate-limiting step for initiation of mammary luminal tumors. The Ink4c-null mouse represents a unique model for study and for developing therapeutic strategies to treat luminal tumors.
Mammary epithelia are comprised of two major epithelial cell types: ductal and alveolar luminal cells, constituting the inner layer of ducts and the lobuloalveolar units, respectively; and a basal layer of myoepithelial cells surrounding the luminal cells. Both epithelial cell types are believed to originate from a common multipotential stem cell (Shackleton et al., 2006; Stingl et al., 2006). Histopathologically, breast cancer can be divided into two major types: luminal tumors that account for the majority of breast cancer cases (70–80%) and are typically estrogen receptor (ER) positive; and basal-like breast cancers that have high histological grades and mitotic indices, are ER-negative, and are associated with a poor prognosis (Althuis et al., 2004). Gene expression profiling analyses have further categorized human breast tumors into several intrinsic subtypes that correlate with tumor biology and are prognostic and predictive for patient outcomes (Hu et al., 2006; Perou et al., 2000; Sorlie et al., 2003). The genetic, cellular, and biochemical mechanisms controlling lineage commitment and maturation in the mammary gland bear clinical significance, as different types of breast tumors possess different outcomes and are diverse in their responsiveness to treatment but remain to be defined. The cellular origin of breast cancer and how its various subtypes arise are not fully understood, and appropriate animal models for studying this disease are still needed.
p18Ink4c (referred to as p18 hereafter) is a member of the mammalian INK4 family that inhibits CDK4 and CDK6, whose activation by mitogen-induced D-type cyclins leads to phosphorylation and functional inactivation of RB, p107, and p130. Functional inactivation of this pathway is a common event in the development of most types of cancer (Sherr, 1996). p18 is a haploinsufficient tumor suppressor in mice whose deletion promotes the development of various tumors, including lymphoma, medulloblastoma, glioblastoma, and tumors of neuroendocrine organs, lung, and prostate (Bai et al., 2003; Bai et al., 2006; Franklin et al., 1998; Latres et al., 2000; Pei et al., 2007; Uziel et al., 2005). Deletion or reduced expression of the p18 gene has also been observed at high frequency in different types of human cancer.
Three lines of evidence suggest a potential function of p18 in suppressing mammary tumors. First, the cyclin D1 gene, located on human chromosome 11q13 and a functional antagonist of p18, is amplified in a substantial portion of human breast cancer [reviewed in (Ormandy et al., 2003)]. Transgenic expression of MMTV-cyclin D1 promotes mammary tumor development (Wang et al., 1994). Second, deletion in mice of cyclin D1 or CDK4, the target of p18, conversely retards mammary gland development and prevents MMTV-oncogene induced mammary cancer (Fantl et al., 1995; Sicinski et al., 1995; Yu et al., 2001; Yu et al., 2006). Lastly, female p18 null mice (in BL/6 background) display mammary gland hyperplasia with ectasia in their galactophor ducts (Latres et al., 2000 and our unpublished observation). We carried out this study to determine the function and mechanism of p18 in suppressing mammary tumorigenesis.
To determine the function of p18 in mammary tumor suppression, we introduced p18−/− into the Balb/c strain, which is known to be susceptible to mammary tumor development and is widely used for such study, by backcross (Ponnaiya et al., 1997; Ullrich et al., 1996). The tumor spectrum, incidence and onset time of many tissues, including pituitary, thyroid, testis, pancreas, and lung, increased in the Balb/c background compared with the BL/6 background (data not shown). The most striking phenotype in Balb/c-p18 mice was mammary gland tumors, as the mammary tumor-free survival was reduced from a mean age of 27.3 months in wild-types (WT), to 20.9 months in p18+/−, and to 15.4 months in p18−/− (Figure 1A).
Mammary tumors developed in Balb/c-p18−/− mice at a very high incidence, including 23% (N=13) of females examined less than one year of age and 88% (N=17, including 1 male) between one year and 20 months (Figure 1B). Most mammary tumors from p18−/− and p18+/− mice were preinvasive carcinomas, such as ductal carcinoma in situ (DCIS), and exhibited papillary, cribriform. and cystic characteristics with low and intermediate nuclear grades (Figure 1C). Only two of 18 (11%) p18−/− mammary tumors were invasive ductal carcinoma (IDC), and one of them metastasized to the lung.
Normal breast ducts contain two types of epithelial cells: luminal cells that are round or polygonal and line the inner ductal lumen; and basal/myoepithelial cells that are elongated and contact the basement membrane but never contact the lumen. Luminal and basal/myoepithelial cells can also be distinguished by their different cytokeratin (CK) expression patterns. The majority of luminal cells express CK7, 8, 18, 19, and 20, while basal/myoepithelial cells typically express CK5, 13, 14, and 17. p18−/− mammary carcinomas, most of which retained an organized gland and duct structure, stained strongly and uniformly for CK8 but not CK5, indicating that they predominantly contain luminal epithelial cells (Figure 1D).
Most (13 out of 15, 87%) p18−/− mammary tumors examined were positive for estrogen receptor alpha (ERα). The percentages of ERα positive cells varied among individual p18−/− mammary tumors, from as low as 2–5% in six tumors, to more than 20% in the other nine, indicating heterogeneity and potentially different stages of tumor development. ERα positive cells were frequently detected in p18−/− mammary tumors, where tumors were well-differentiated and still retained an overall normal architecture of mammary glands (Figure 1E, left and middle), or when poorly-differentiated tumors were investigated (Figure 1E, right). When compared with 13 different murine mammary tumor models (Herschkowitz et al., 2007), p18−/− mammary tumors express the highest level of ER (data not shown). It should be pointed out, however, that most mammary tumors developed in p18−/− mice are DCIS, with only 11% IDC or metastatic tumors, whereas mammary tumors developed in other murine models are mostly invasive carcinoma. It remains to be determined whether the difference in ER expression between p18−/− tumors and tumors of 13 other models reflects different stages in tumor progression or inherent differences in tumor biology. Together, these results suggest that p18 plays an important role to suppress the development of ERα-positive luminal mammary tumors in mice.
We determined cell proliferation in mammary tumors developed in p18−/− mice (>14 months of age) by examining both the proliferation marker Ki67 (Figure 2A) and the mitotic marker phosphorylated histone H3 (data not shown). Both analyses revealed that there was active cell proliferation in p18−/− mammary tumors compared with mammary tissues derived from age-matched p18+/− or WT mice. As determined microscopically by their inner localization in the mammary glands, a majority of proliferating cells in p18−/− mammary tumors were luminal cells.
We then determined mammary epithelial proliferation by pulse-labeling 9-month-old virgin littermate WT and p18−/− mice with BrdU. At this age, most p18−/− mammaries exhibit either normal appearance or hyperplasia and retain relatively normal architecture of mammary glands but not yet developing tumors. There was very little cell proliferation in WT ductal epithelium, and as expected, BrdU incorporation was barely detectable (Figure 2B, note that green fluorescence outside the glands does not correspond to individual cells and was likely caused by non-specific staining). In contrast, BrdU-positive cells in littermate p18−/− ductal epithelium were substantially increased. Microscopic examination of four independent mammary glands from three different mice of each genotype revealed a 3.8-fold increase of BrdU positive cells in p18−/− mammary glands (6.4% ± 2.3%) than in WT glands (1.7% ± 0.7%, Figure 2B). Most p18−/− ducts/glands contained more than two BrdU+ cells, and nearly all BrdU+ cells were also positive to CK8 staining, indicating that the majority of BrdU+ cells were luminal epithelia. Only a few BrdU-positive cells co-expressed CK5, a marker for myoepithelial cells (data not shown).
At eight weeks of age, mammary ducts branch with active epithelial cell proliferation, including both luminal and myoepithelial cells, and no lesions were detected in p18−/− mammaries as confirmed by parallel H&E staining (data not shown). p18 deficiency resulted in a 2.5-fold increase of BrdU positive cells in mammary epithelium, most of which were luminal-like epithelial cells (Figure 2C). Finally, we directly examined the status of Ser608 phosphorylation of Rb protein, the site that is preferentially phosphorylated by p18’s targets CDK4 and CDK6 (Schmitz et al., 2004; Zarkowska et al., 1997). A visible and consistent increase of pRb-Ser608 phosphorylation, both in the intensity and number of positive cells, was seen in normal p18−/− mammary epithelia (11.2 ± 1.8% in wt to 23.5 ± 3.8% in p18−/−) and p18−/− tumors (34.5 ± 5.7%, Figure 2D), supporting the activation of CDK4 and/or CDK6. Most Rb-Ser608 positive cells were luminal cells. Taken together, these results demonstrate that p18 deficiency stimulates CDK4 and/or CDK6 activity towards the Rb protein in luminal epithelial cells, increasing their proliferation in a cell autonomous manner at an early age prior to the development of any detectable preneoplastic lesion. Loss of p18 also maintains luminal cells at a hyperproliferative state throughout adulthood, during which time p18−/− luminal cells progressively develop from hyperplasia to tumor.
We explored the possibility that p18 loss stimulates mammary stem and/or luminal progenitor cell proliferation. One unique feature of stem/progenitor cells is their slow rate of proliferation. As a result, the BrdU label incorporated during a long period of pulse labeling can be retained after a long period of chase by the stem/progenitor cells but are continuously diluted in other somatic cells by each cell cycle. Label-retaining cells (LRCs) have been demonstrated to be enriched in both mammary stem and progenitor cells (Shackleton et al., 2006; Smith, 2005; Welm et al., 2002). We injected BrdU into pubertal littermates twice a day consecutively for seven days and determined the LRCs after short (1 day) and long (37 days) chase. At day 1 of chase, more than half of the epithelial cells were BrdU positive in both WT (51.7% ± 6.8%) and p18−/− mammaries (55.3% ± 8.1%) (Figure 3A). After 37 days of chase, BrdU positive LRCs were only sparsely detected in the WT mammaries but were notably increased in p18−/− mammaries (Figure 3B). Microscopic examination of 12 mammary glands from three mice of each genotype revealed a 1.6-fold increase of LRCs in p18−/− mammary glands (14.5% ± 1.8%) over WT glands (9.1% ± 1.3%).
Most LRCs in both WT and p18−/− mammaries were round or polygonal, located in the inner layer of the duct, and made contact with the ductal lumen, indicative of luminal epithelium. To confirm this, we stained mammary sections with CK8 and CK5 and determined that indeed, most WT (7.3% ± 1.4% of total mammary epithelium and 80.2% of total LRC) and p18−/− LRCs (12.04% ± 1.7% of total mammary epithelium and 83% of total LRC) were positive for CK8 (Figure 3B). Notably, the CK8-positive LRCs, which are likely enriched for luminal progenitor cells, were increased by 1.7-fold by p18 loss, from 7.3% of total epithelial cells in WT mammaries to 12% in p18−/− mammaries (Figure 3B).
Previously, Smith and colleagues combined light and electronic microscopic analyses of the mammaries from different mammals and defined five histologically distinct epithelial cell populations based on the cell size and nuclear and cytoplasmic staining characteristics: small light cell (SLC), undifferentiated large light cell (ULLC), differentiated large light cell (DLLC), large dark cell (LDC), and myoepithelial cell (MYO) (Chepko et al., 2005; Smith and Medina, 1988). Both small and large light cells appear to be division-competent, based on their condensed chromosomes and capability to undergo mitosis in explant culture. Combined with the characterization of the cytoplasmic organelle differentiation and localization in the glands, these features led to the proposal that SLCs represent a population enriched for mammary stem cells that divide and differentiate progressively into ULLC, likely corresponding to the luminal progenitors, and then to DLLC, before terminal differentiation into a bulk population of mature luminal cells (LDCs).
To determine the division-competency of these structurally defined cell populations and their label retaining ability, we took immunostained sections from the label retaining assay (37-day chase), re-stained them with H&E, and matched 793 cells from WT and 946 cells from p18−/− mammaries from two stainings by side-by-side microscopic comparison (Figure 3C and Table S1). Of these matched cells, 70 from WT and 150 from p18−/− mammaries were LRCs and close to half, 44% in WT and 48% in p18−/− mammary, respectively, corresponded to ULLCs. The remaining LRCs were similarly distributed among SLCs, DLLCs, LDCs, and MYOs (Table S1). This result indicates that ULLCs represent the majority of LRCs, providing functional support to the histologically defined notion that ULLCs represent a progenitor-enriched mammary epithelial cell population.
We next microscopically examined more than 20 ducts and lobules from each animal of both genotypes between two and three months of age, and counted at least three lobules and one duct per animal (Figure S1). Of five major populations, the ULLC population was significantly increased from 16.5% in ducts and 19% in lobules of WT mammaries, to 28.1% and 25% in p18−/− ducts and lobules, respectively, and led to a 50% overall increase of ULLCs from WT (17.6%) to p18−/− (26.5%) mammaries (Figure 3D, and Figure S1). The SLC population was very small (less than 3%) and partially overlapped with myoepithelial cells in their localization, making it difficult to determine whether p18 deletion also affects the size of the putative mammary stem cell population. There is no significant difference in relative distribution of the other three populations, as well as cells of unknown (UK) origin, between WT and p18−/− mammaries (Figure S1). These results demonstrate that luminal progenitor-enriched ULLCs are the major cell population affected by the loss of p18 at an early age, prior to tumor development.
To corroborate the finding that loss of p18 expands luminal progenitor cells, we dissected mammary glands from 6–8 week mice and analyzed them by flow-cytometry. After exclusion of lymphocytes (SSC low), dead cells (7AAD positive), hematopoietic cells (CD45+TER119+), and endothelial cells (CD31+), the expression profile for CD24 and CD29 was determined (Figure S2A). The CD29loCD24+ population, that is enriched with progenitors and have a luminal cell fate under lactogenic condition (Asselin-Labat et al., 2007; Shackleton et al., 2006; Sleeman et al., 2007), was increased by 1.5-fold in p18−/− mammaries compared to WT littermates in Balb/C (42.3 ± 2.5% vs 27.9 ± 3.7%, Figure 4A) and 1.4-fold in BL/6 (51% vs 35.9%, Figure S2B) backgrounds at pubertal age (6–8 week). The CD29loCD24+ cells were very similar between p16−/− and p16+/− (26.52% vs. 25.83%, data not shown), supporting a specific role of p18 in controlling luminal progenitor cell proliferation in pubertal mice. The CD29loCD24+ cell population was similarly increased in postpubertal (10–12 week) p18−/− mammaries (42.7 ± 4.0% vs 34.1 ± 2.6%, Figure 4A), and remained high late in life in p18−/− mice, 48.3% at 9 months of age and 37.8% (in hyperplasic tissues) or 42.1% (in tumorigenic tissues) at 14 months of age, and both were higher than WT littermates (41% and 34% at 9 and 14 months of age, respectively). These data suggest that p18 loss results in a continuous increase of luminal progenitor-enriched CD29loCD24+ cell population during aging and development.
CD49f (α6 integrin) is expressed in human mammary colony forming cells (Ma-CFC) enriched for luminal progenitor cells and has also been used in combination with CD24 to identify mouse Ma-CFCs and stem cells (Stingl et al., 2001, Stingl, 2006). After exclusion of lymphocytes and dead, hematopoietic, and endothelial cells, we determined that the CD24highCD49flow population was significantly increased in p18−/− mammaries compared with WT littlemates by 1.8-fold in Balb/c (6 wk, 7.9% vs 4.2%, Figure 4B) and 1.7-fold in BL6 (12 wk, 6.6% vs 3.8%, Figure S2D) backgrounds. The CD24highCD49fhigh population enriched for mammary stem cells, on the other hand, did not appear to be significantly affected by p18 loss (3.6% vs. 3.2%, Figure 4B). The CD24highCD49flow population in p18−/− mammary at 11 months of age also remained higher than in WT littermate mammaries (5.5% vs 4.0%, Figure S2E). These results further support the function of p18 in constraining luminal progenitor expansion.
CD24+ mammary epithelial cells are enriched for luminal cells and can be further separated into two distinct populations based on the expression of either Sca1 or prominin-1, which almost completely overlap: a CD24+Sca1− or CD24+prominin-1− population that is enriched (>40%) for colony-forming cells (CFCs, i.e. luminal progenitor cells); and a CD24+Sca1+ or CD24+prominin-1+ population that is enriched (>80%) for cells expressing genes involved in sensing systemic hormones such as ERα, progesterone receptor, and prolactin receptor and contains little stem cell activity (Sleeman et al., 2006; Sleeman et al., 2007). Consistent with the expansion of the CD29loCD24+ cell population, the CD24+Sca1− population was also higher in p18−/− mice than in littermate WT mice at postpubertal age (28.3% vs.23.3%) and maintained at a higher level later in life; 59% in p18−/− vs. 50% in WT mice at 9 months of age and 57% in p18−/− vs. 41.7% in p18+/− mice at 14.5 months of age (Figure 4C). Similar results, that p18 deficiency increased both CD24+Sca1− (14% vs. 9%) and CD24+Sca1+ populations (39% vs. 25%) when compared with WT littermates, were also obtained in BL/6 mice (Figure S2C), indicating that the function of p18 in constraining luminal cells is not strain-dependent.
To directly demonstrate that the proliferation of luminal progenitor cell populations is increased in p18−/− pubertal mice, we isolated mammary epithelial cells and stained them with CD24 and intracellularly with the proliferation marker Ki67 after depletion for lymphocytes and Lin+ cells. The CD24+Ki67high population was increased from 29.3% in WT mammaries to 43.1% in p18−/− mammaries (Figure 4D), providing further evidence to conclude that p18 loss initiates the expansion of a luminal progenitor-enriched cell population.
We next determined the function of p18 in constraining luminal progenitor cell proliferation using three different in vitro systems: 3D Matrigel, mammosphere, and colony formation assays. The Matrigel assay showed that mammary cells derived from p18−/− mice exhibited a substantial increase of proliferation and that the majority of BrdU positive cells were luminal (Figure 5A), demonstrating that the increased proliferation of luminal cells in p18−/− mice is cell autonomous. Mammary stem/progenitors propagated in nonadherent culture in vitro form mammospheres and are able to differentiate into both luminal and basal/myoepithelial lineages (Dontu et al., 2003). Cells derived from p18−/− mammaries showed an increase in primary mammosphere forming activity, 32.8 ± 5.1 spheres/20,000 cells, than WT cells, 26.3 ± 3.4 spheres/20,000 cells (p=0.027, paired t-test, Figure 5B), indicating increased stem or progenitor cell function resulting from p18 loss. Most WT spheres were 70–100 μm in size with 10–20% larger than 100 μm, whereas most p18−/− spheres were 80–120 μm in size with 10–20% spheres larger than 120 μm (Figure 5B). The average size of p18−/− spheres (118 ± 38 μm) was significantly larger than WT spheres (92 ± 24 μm, p=0.017, paired t-test, Figure 5B), further supporting the conclusion that p18 loss increases the proliferative potential of mammary progenitor cells. To examine their differentiation potential, we induced individual mammospheres to differentiate. After four days of culture in gelatin-coated dish, most cells in the spheres were attached, differentiated, and migrated out of the center. Immunostaining identified CK8+ and CK5+ cells in both WT and p18−/− spheres (Figure S3), confirming that the spheres were capable of differentiating into both luminal and basal/myoepithelial lineages and that p18 deficiency did not affect the differentiation potential of the mammosphere.
We then compared the colony-forming capacity of WT and p18−/− mammary epithelial cells (MECs) in Matrigel and scored both spherical and flat colonies. For scoring spherical colonies, we divided them into >50 μm and >100 μm groups and found that p18−/− MECs formed more spherical colonies in both groups, averaging 10.8 colonies >100 μm or 9.5 colonies >50 μm per 1,000 MECs seeded, compared to WT or heterozygous MECs that formed on average 6.8 colonies >100 μm or 4.5 colonies >50 μm per 1,000 MECs seeded (Figure 5C). The p18−/− MECs also appeared to form more flat colonies in both size groups, but the limited number of flat colonies formed by both WT and p18−/− MECs prevented us from making a statistically significant determination (Figure 5C). Because we used MECs in these FACS and colony-formation assays, we cannot exclude the possibility that p18 may also have a function in regulating stroma cells in vivo, which could contribute to the control of luminal progenitor cell expansion. Almost all spherical colonies generated from either WT or p18−/− MECs, including both acini with or without lumen, stained strongly for CK8, with varying degrees of CK5 staining (Figure S4A). Staining of colonies after dissolving Matrigel revealed two distinct groups of colonies: one containing cells in the center that were strongly stained for CK8 and a few outer layer cells that were stained positively for CK5 (Figure S4B, left panels); and one comprised uniformly of CK8+ cells throughout the colony (right panels).
To obtain a quantitative assessment, we generated more colonies by seeding a high density of unsorted mammary cells on Matrigel (20-fold more than clonal density). Flat colonies were difficult to identify accurately due to their high density and were not counted. p18−/− mammary cells formed significantly more spherical colonies in both >50 μm and >100 μm size groups and slightly more colonies larger than 200 μm compared with WT littermate mammary cells (Figure S4C). Most spherical colonies stained strongly for CK8 with varying degrees of CK5 staining.
Considering that serum-free culture conditions favor luminal cell proliferation, which may cause an underestimation of the myoepithelial cells, we divided the spherical colonies into two groups: ‘basal colonies’ (CK8+ and more than 50% CK5+ cells) and ‘luminal colonies’ (CK8+ and less than 50% CK5+ cells) (Figure 5D). Microscopic examination demonstrated that p18−/− mammary cells generated 1.6-fold more luminal colonies (377± 50.3/20,000 cell seeded) than did WT mammary cells (232±30.1/20,000 cell seeded, Figure 5D). p18−/− mammary cells generated slightly more CK5+ cell enriched basal-like colonies (104±14.7/20,000 cells seeded) than the WT MECs (87±17.3/20,000 cells seeded), but the increase was not statistically significant (p=0.27, Figure 5D). This result supports a unique function of p18 in luminal progenitor proliferation and differentiation.
To determine whether our mouse genetic analysis models human breast cancers, we queried the gene expression data of the Netherlands Cancer Institute 295 (NKI295) breast cancer patient sample set (van de Vijver et al., 2002). The 295 patients were split 50/50 into a “high” and “low” expression group for each analyzed gene and the distribution of ‘high’ vs. ‘low’ expression of each gene was determined relative to five intrinsic tumor subtypes: basal-like, Her2+/ERα−, Luminal A, Luminal B and normal-like as assigned (Fan et al., 2006). Using multiple statistical tests, we determined that p18 mRNA expression levels were highly correlated with the breast tumor intrinsic subtype (chi-squared p<0.0001 (Figure 6A, and Table S2)). Notably, p18 as well as p19 INK4D (data not shown) levels, but not five other CDK inhibitor genes, tended to be low in the Luminal A subtype of human breast cancers (p = 0.003), which are almost exclusively ERα+ and high in basal-like tumors (p <0.0001). These results are consistent with the finding that mammary tumors developed in p18−/− mice are predominantly comprised of luminal cells and are ERα positive. Interestingly, RB1 expression is low in basal-like tumors (p=0.0001) and high in luminal A tumors (p=0.0008) (Figure 6A, and Table S2). An inverse correlation between p18 and RB1 expression is consistent with the notion that p18, among four INK4 genes, is the major upstream activator of RB1, and its decrease may relieve selection pressure to decrease or mutate RB1 during tumorigenesis in mammary luminal cells.
GATA3, a member of the GATA family of Zinc finger transcriptional factors, was recently identified to play an essential function in differentiation and maintenance of luminal cell fate and thus for mammary-gland morphogenesis. Restoration of GATA3 in late carcinomas from MMTV-PyMT mice induced tumor differentiation and suppressed tumor dissemination (Kouros-Mehr et al., 2008). Conspicuously, the expression of GATA3 is high in luminal A and low to absent in basal-like tumors, thus exhibiting an inverse correlation with p18 (Figure 6A, and Table S2). Scatter plot analysis further confirmed the inverse correlation between GATA3 and p18 mRNA levels in human breast cancer patients from the NKI patient series (Correlation coefficient = −0.379, Figure 6B). This finding suggests the possibility of negative regulation of p18 by GATA3 (see below).
We next examined whether the levels of p18 expression is related to patient outcomes and included all CDK inhibitor genes in the analysis to explore the specificity. Kaplan-Meier analysis of overall survival revealed that “high” vs. “low” expression of p18INK4C, p19INK4D, and RB1 genes, but not p21CIP1/WAF1, p27KIP1, p57KIP2, or p16INK4A, was significantly predictive of patient outcomes. High INK4C and INK4D expression and low RB1 expression predicted poor patient outcome (Figure 6C). Identical results were obtained for relapse-free survival (Figure S5). The statistically significant value of p18 levels in predicting patient outcome is consistent with, and can be explained by, the correlation between low p18 expression and luminal A subtype tumors exhibiting a better outcome (Hu et al., 2006). The profiling analyses linking the expression of INK4 and RB1 genes with patient outcomes also suggest how an impairment of the RB pathway may clinically predict, and mechanically influence, patient outcomes. A reduction of RB1 level, among all genes acting on the RB pathway, would be most direct and effective in diminishing the activity of the RB pathway and promoting cell proliferation, and is therefore linked with, or leads to, a more malignant phenotype (i.e. basal-like tumors), while a reduction of its individual upstream activators, such as p18INK4C, is less potent in promoting cell division and is linked with less aggressive tumors (luminal A tumors). Notably, contrary to p18 expression, high GATA3 expression predicts a better patient outcome, thus showing that human luminal A tumors are characterized by high GATA3, high RB1, and low p18.
Prompted by the inverse correlation between p18 and GATA3 expression in human luminal A breast cancer patients, and the recent evidence that Gata3 is a highly enriched transcription factor in the mammary epithelium of pubertal mice and plays a critical role in luminal cell differentiation (Asselin-Labat et al., 2007; Kouros-Mehr et al., 2006), we hypothesized that p18 may be a downstream target of GATA3 in controlling luminal cell proliferation and differentiation. We first examined the expression of Gata3 and p18 in different populations of mammary epithelium. Sorted luminal progenitor enriched C24+CD29− cells from WT mice at a pubertal age (Shackleton, 2006) expressed 12-fold more Gata3 than C24−CD29− cells that include basal/myoepithelial cells and other cell types (Figure 7A), confirming that Gata3 is predominantly expressed in luminal epithelium. Interestingly, the Gata3 mRNA level in MaSC-enriched C24+CD29+ cells was also much higher than in C24−CD29− cells, suggesting a potential, and yet to be defined, function of Gata3 in MaSCs. Again, we observed that the p18 mRNA level was inversely correlated with Gata3; it was significantly lower in both MaSC-enriched C24+CD29+ and luminal progenitor enriched C24+CD29− cells than in C24−CD29− cells (Figure 7A).
To directly determine whether Gata3 negatively regulates p18 in mammary epithelium, we first determined the p18 protein level in Gata3-deficient mouse mammary tissue. Specific deletion/reduction of Gata3 in the mammary gland was achieved by pregnancy of Gata3f/f;WAP-cre dams and resulted in a robust increase of p18 protein (Figure 7B). Likewise, the knocking down of GATA3 expression in human MCF-7 and 293T cells by two different siRNA oligonucleotide duplexes resulted in increased p18 mRNA and protein levels (Figures 7C, 7D). Conversely, overexpression of GATA3 in MCF-10A cells inhibited p18 mRNA expression (Figure 7E) and increased cell proliferation, as determined by BrdU incorporation (Figure 7F). We also examined Gata3 expression both in vivo in animals at different ages and in vitro in acini cultured in 3D Matrigel using mammary cells. Gata3 expression was readily detected, but no discernible difference of Gata3 expression was detected between the WT and p18−/− mammary epithelial cells (data not shown).
Examination of the p18 promoter reveals several GATA3 consensus sites, including a 40-bp region containing 10 GATA3 sequences in tandem located 5.3 kb upstream of the translational initiation ATG codon (Figure 7G). To determine if GATA3 directly binds to the p18 locus, we analyzed 6.6 kb upstream and 2.3 kb downstream of the start codon of the human p18 gene by chromatin-immunoprecipitation (ChIP) assay using 33 pairs of primers. Six out of 33 amplicons, four upstream (a, b, c, and d in Figure 7G) and two downstream of the start codon (h and i in Figure 7G), were specifically enriched in the GATA3 immunoprecipitation (Figure 7G). Except for amplicon h, all positive amplicons contain at least one putative GATA3 binding site (data not shown), including one that contains 10 GATA3 repeats. Together, these results demonstrate that GATA3 binds to the p18 gene and negatively regulates its transcription.
In this study, we report that nearly 90% of female p18 null mice in the Balb/c background spontaneously developed mammary tumors with a median tumor-free survival time of 14 months. To the best of our knowledge, Balb/c-p18 mice develop mammary tumors at the highest rate among all mutant mice with germline mutations. Our findings support the rate-limiting function of INK4-cyclin Ds-CDK4/6 in suppressing mammary tumorigenesis.
The results presented here also suggest the existence of a genetic modifier(s) in Balb/c mice that interacts with p18-deficiency epistatically to promote tumorigenesis. The possibility of p16Ink4a, a tumor suppressor that sustained two allelic variations in Balb/c affecting its inhibitory function toward CDK4 (Zhang et al., 1998), has been ruled out by our observation that p16−/−;p18−/− in CJ57BL/6 background did not develop mammary tumor at an elevated rate (Ramsey et al., 2007). Identifying the modifier(s) remains a daunting challenge, especially when considering multiple modifiers could have contributed to the tumor susceptibility phenotype in Balb/c.
The mammary tumors developed in Balb/c-p18 mice are predominantly luminal. In addition to data presented here, gene expression profiling analysis clustered most tumors derived from p18−/− mice with other mice known to have luminal mammary tumors (e.g. MMTV-Neu and MMTV-PyMT) and showed expression of the XBP1 cluster (unpublished results). Our studies demonstrate that loss of p18 selectively stimulates mammary luminal progenitor cell proliferation at an early age and maintains them at a hyperproliferative state throughout life, leading to increased and sustained proliferation of luminal epithelial cells and eventual development of luminal tumors. Furthermore, the mammary tumors developed in p18−/− mice are ER+. These features establish the Balb/c-p18 mouse as an animal model for elucidating the molecular mechanism and cell origin of luminal breast cancers and for preclinical studies of treatment for ER+ luminal tumors.
GATA3 is expressed early during embryogenesis and specifically in both ductal and alveolar luminal epithelial cells. Genetic analyses have recently demonstrated that Gata3 plays multiple morphological and functional roles in mammary development, including terminal end bud, placode and nipple formation, ductal elongation and invasion, and subsequent lactation (Asselin-Labat et al., 2007; Kouros-Mehr et al., 2006). Blockage of luminal progenitor cells to differentiate was recognized as the most likely cellular basis underlying these various functions; the luminal progenitor pool was increased with a concurrent decrease of differentiated luminal cells in MMTV–Cre;Gata3f/f mice, while ectopic expression of Gata3 in MaSC-enriched cells promoted the stem cells to differentiate along the alveolar cell lineage. Loss of p18 resulted in expansion of luminal progenitor cells, suggesting that p18 functions to either restrict stem cells from asymmetric divisions and produce luminal progenitors or to maintain luminal progenitors at a quiescent state (Figure 7H). We further suggest that GATA3 promotes the differentiation of mammary stem cells along the luminal lineage through a network of transcription regulation that, on one hand, represses p18 expression to allow the expansion of either luminal progenitors or transit cells of luminal lineage and, on the other hand, activates the expression of genes such as FOXA1 and ERα to induce and maintain the differentiation of mature luminal cells. Further investigations are necessary to determine the factor that trans-activates and/or maintains p18 expression in luminal progenitors.
Loss of Gata3 would result in an accumulation of p18 in the luminal progenitor cells, blocking them from entering an active cell cycle and undergoing subsequent differentiation. Loss of p18, although stimulating the proliferation of luminal progenitor cells and initiating luminal tumorigenesis, would not significantly impair the differentiation of luminal lineage in the presence of normal GATA3 function. Our study also provides a plausible molecular explanation underlying the pathological association and prognostic feature of GATA3. A high level of GATA3, which has been associated with less aggressive lower grade cancers (Mehra et al., 2005; Sorlie et al., 2003; van de Rijn et al., 2002), would suppress p18 expression, and a decreased p18 level would favor the development of luminal type breast cancers that have a better outcome for patients.
Genetic studies have recently linked the function of p16 with the control of stem and progenitor cells during aging in neuronal, hematopoietic, and pancreatic lineages (Janzen et al., 2006; Krishnamurthy et al., 2006; Molofsky et al., 2006). These genetic analyses, combined with three lines of evidence from biochemical properties, structural analyses, and the patterns of gene expression, support a model that p16 and p18 collaboratively control stem and progenitor cell cycles. First, rapid turnover of the p21 family of CDK inhibitors and intrinsic stability of INK4 proteins make p21 and INK4 family proteins more suitable for inducing and maintaining a transient and stable cell cycle arrest, respectively. Second, individual INK4 proteins bind to CDK4 and CDK6 with similar affinity. As a result, the activity of CDK4/6 in a given cell is determined by the total concentration of all INK4 combined, rather than individual INK4, making them functionally related and collaborative. Thirdly, p18 is expressed early and sustains a high level throughout life in many adult tissues, whereas the expression of p16 is undetectable in young tissues and is induced during aging (Krishnamurthy et al., 2004; Zindy et al., 1997). These observations led us to propose that p18 and p16 collaboratively constrain stem cell self-renewal and progenitor cell proliferation, with p18 playing a major role in maintaining the homeostasis of stem/progenitor cells from early embryogenesis throughout adulthood and p16 in limiting stem/progenitor cell function during aging.
The generation and genotyping of p18 and floxed Gata3 mice have been described previously (Franklin et al., 1998; Pai et al., 2003). p18 mutant mice in BL/6 background were backcrossed for seven generations with BALB/c mice before used for this study as Balb/c-p18 mutant mice. Tissues of most organs were removed, fixed in 10% neutral buffered formalin, and examined histologically by two pathologists after H&E staining. Immunohistochemistry and primary antibodies were as described previously (Bai et al., 2003). The IACUC (Institutional Animal Care and Use Committee) at UNC approved all procedures.
To examine and quantify small light cell (SLC), undifferentiated large light cell (ULLC), differentiated large light cell (DLLC), large dark cell (LDC), and myoepithelial cell (MYO), we followed criteria defined by Smith and colleagues (Chepko et al., 2005; Smith and Medina, 1988). Cells that could not be clearly defined were categorized as unknown (UK). 3 pairs of WT or p18−/− mice at 2–3 months of age were dissected and stained with H&E. All sections were read by two different pathologists (X-H.P. and B. F.). Only sections containing large amounts of epithelium of sufficient quality for counting were selected. Epithelial structures were classified as ducts when they were singular elements associated with no other epithelial structures. They were classified as lobules if they appeared clustered with other small, round-shaped epithelial structures or if there were small round-shaped structures that budded from a larger, longitudinally sectioned duct. Cells in terminal end buds were also included in the lobule count.
For short-term in vivo BrdU labeling analysis, mice were injected with 1ml per 100g body weight BrdU cell proliferation labeling reagent (Amersham Bioscience, Little Chalfond, UK) one hour before sacrificing. Tissues were immunostained with anti-BrdU antibody (Abcam, Cambridge, MA). To determine label retaining cells (LRCs), littermate mice at one month of age were injected intraperitoneally with BrdU (50mg/kg) twice a day for seven consecutive days followed by chase for one day or 37 days before sacrificing. 4 μm mammary sections were immunostained with antibodies against BrdU, CK5, and CK8. For counting LRCs, only sections containing a large amount of mammary epithelium of sufficient quality were selected, and only confirmed BrdU positive cells in the duct were counted. Parallel staining of the serial sections for CK8, BrdU, and DAPI was performed to calculate the percentage of CK8+BrdU+ cells in DAPI+ mammary epithelium.
Four mammary glands (the 4th and 5th pairs) from each mouse were examined, and at least 600 cells per mammary (more than 2400 epithelial cells per mouse) were counted in the parallel staining of serial sections. The average percentages of BrdU+ cells were calculated from three WT and three p18 null mice, respectively.
We used a data set of breast-cancer samples from 295 women (van de Vijver et al., 2002). Patients were dichotomized into two groups as above or below median for each gene’s expression value (2 based log ratio of red channel intensity and green channel intensity). Kaplan-Meier survival plots were performed for Relapse-Free Survival (RFS) and Overall Survival (OS). Cox-Mantel log-rank tests were applied to compare the survival differences and obtained p-values, using WinSTAT for Excel (R. Fitch Software). We then compared the two groups of patients for each gene versus five breast cancer subtypes(Hu et al., 2006; Sorlie et al., 2003) using two-way contingency-table analyses and SAS software (Cary, NC). All of the statistical analyses were performed with StatsDirect 2.4.3 software (StatsDirect Statistical Software). The survival rate was calculated by the Kaplan-Meier method.
Detailed descriptions for these experimental procedures are available in the Supplemental Data.
We thank Yojiro Kotake for helping with ChIP assay and Ned Sharpless, Chuxia Deng and Connie Eaves for discussions. F. B. was supported in part by a U.S. Department of Defense Career Postdoctoral fellowship. This study was supported by the NCI Breast SPORE program (P50-CA58223) and the Breast Cancer Research Foundation grants to C.M.P, and an NIH grant (CA68377) to Y.X.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.