Dynamic role of the codon 72 p53 single-nucleotide polymorphism in mammary tumorigenesis in a humanized mouse model

Female breast cancer (BrCa) is the most common noncutaneous cancer among women in the United States. Human epidemiological studies reveal that a p53 single-nucleotide polymorphism (SNP) at codon 72, encoding proline (P72) or arginine (R72), is associated with differential risk of several cancers, including BrCa. However, the molecular mechanisms by which these variants affect mammary tumorigenesis remain unresolved. To investigate the effects of this polymorphism on susceptibility to mammary cancer, we used a humanized p53 mouse model, homozygous for either P72 or R72. Our studies revealed that R72 mice had a significantly higher mammary tumor incidence and reduced latency in both DMBA-induced and MMTV-Erbb2/Neu mouse mammary tumor models compared to P72 mice. Analyses showed that susceptible mammary glands from E-R72 (R72 x MMTV-Erbb2/Neu) mice developed a senescence-associated secretory phenotype (SASP) with influx of proinflammatory macrophages, ultimately resulting in chronic, protumorigenic inflammation. Mammary tumors arising in E-R72 mice also had an increased influx of tumor-associated macrophages, contributing to angiogenesis and elevated tumor growth rates. These results demonstrate that the p53 R72 variant increased susceptibility to mammary tumorigenesis through chronic inflammation.


Introduction
Female breast cancer (BrCa) affects more than 2.2 million women each year worldwide [1]. Etiology of BrCa is heterogeneous and involves genetic, environmental, and lifestyle factors. Only 5-10% of BrCas cases are hereditary [2], indicating that the majority are due to genetic susceptibilities affecting response to environmental exposures.
The tumor suppressor p53 is the most commonly mutated gene in BrCa, and functional disruption of p53 is associated with increased tumor aggressiveness, refractoriness to treatment and poorer prognosis [6,7]. P53 regulates cell-autonomous biological activities that maintain genomic integrity, particularly in response to insults such as DNA damage, oncogene activation, and oxidative stress. In response to these stressors, p53 transactivates genes that regulate apoptosis, cell cycle arrest and senescence, facilitating repair or elimination of irreparably damaged cells [8]. However, recent evidence indicates that the anti-cancer effects of p53 can be undermined when persistent cell cycle arrest and accumulation of senescent cells leads to the acquisition of a senescence-associated secretory phenotype (SASP). SASP is characterized by synthesis and secretion of a variety of proinflammatory cytokines and chemokines resulting in chronic inflammation that is protumorigenic [9].
A common single-nucleotide polymorphism of p53 at codon 72 modifies critical biological processes that impact susceptibility to several cancers. This SNP encodes either proline (P72) or arginine (R72), and is located in the polyproline region of exon 4, between the transactivation and DNA binding domains [10,11]. In vitro studies with human cancer cells and rodent fibroblasts show that following DNA damage, the P72 variant preferentially promotes cell cycle arrest [12], senescence [13], and DNA repair [14], while the R72 variant more effectively induces apoptosis [15][16][17].
In vivo analyses in Hupki (Human p53 exon 4-9 knockin) mice have shown that these codon 72 variants induce apoptosis and senescence in a tissue and context-dependent manner [18,19]. Exposure to radiation significantly induces levels of apoptosis in the small intestine of R72 compared to P72 mice. However, P72 animals have significantly higher levels of apoptosis in the thymus [18]. When Hupki mice are fed an obesogenic, high-fat diet, the codon 72 variants differentially regulate genes involved in cellular metabolism and inflammation [19].
Human epidemiological studies of cancer risk have shown modest, but significant associations of codon 72 p53 variants with incidence in several types of cancers, including lung, prostate, and breast [20][21][22][23][24]. Similar to the Hupki mouse, codon 72 p53 variants display cell-and tissuespecific activities in humans, and are not associated with increased risk for all cancer types [11].
The mechanisms by which codon 72 p53 variants differentially affect breast cancer susceptibility remain unresolved. Therefore, we explored this question in a physiologically relevant in vivo setting, using a well-characterized, humanized p53 exon 4 knock-in mouse model, expressing the codon 72 polymorphisms [17]. Our results showed that the R72 variant induced significantly greater levels of chronic mammary tissue inflammation compared to the P72 variant, contributing to enhanced susceptibility to mammary carcinogenesis.

Increased mammary tumor incidence and reduced latency in R72 mice
Initial investigations focused on the impact of the p53 variants at codon 72 on mammary tumor development in two distinct models. First, 7,12-dimethylbenz(a)anthracene (DMBA) was used to induce mammary carcinomas [25]. As shown in Fig. 1A, mice homozygous for the arginine variant, R72 (n = 39), had significantly reduced mammary tumor latency compared to those with the proline variant, P72 (n = 40, Log-rank test; p = 0.049). Mammary tumors first appeared at 21 days post-DMBA in R72 mice, 50 days earlier than in the P72 animals. R72 mice also had a 50% higher mammary tumor incidence compared to P72 animals (Fig. 1B). Mice in the vehicle-treated groups remained tumor-free, indicating that the codon 72 variants did not stimulate spontaneous mammary tumorigenesis in this model (Fig. 1A).
Next, the impact of p53 polymorphic variants on mammary tumorigenesis was investigated in the MMTV-Erbb2/ Neu FVB mouse model [26]. Mice were cross-bred to generate study animals that were homozygous for the codon 72 variants and hemizygous for the MMTV-Erbb2/Neu transgene (E-P72R). As in the carcinogenesis study, E-R72 mice (n = 56) had significantly reduced mammary tumor latency compared to E-P72 (n = 54, Log-rank test, p = 0.004, Fig. 1C). Mammary tumors first appeared in E-R72 animals as early as 122 days of age, compared to 173 days in E-P72 animals, a 51-day difference. Mean tumor latencies of E-R72 and E-P72 mice were 226 and 250 days, respectively. Mammary tumor incidence was 17% higher in E-R72 animals compared to E-P72 animals (p = 0.004, Fig.  1D). As previously observed in MMTV-Erbb2/Neu mice, which are on an FVB background [26], mammary tumors in both E-R72 and E-P72 mice were mammary adenocarcinomas with moderate to poor differentiation (Fig. 1E, F). These results show that in comparison to P72, R72 mice had increased mammary tumor incidence and reduced latency in both carcinogenesis and genetic models. Additionally, these findings demonstrate that the p53 variants altered the susceptibility to tumorigenesis in response to chemical (DMBA) or oncogenic (Erbb2) stimuli.
These latency results prompted an assessment of tumor progression. Tumors in E-R72 animals had a significantly higher average growth rate (37.8 ± 5.1 mm 3 per day) compared to tumors in E-P72 mice (22.8 ± 5.0 mm 3 per day) ( Fig. 2A). Immunolocalization of Ki67 revealed significantly higher levels of proliferation in tumors from E-R72 compared to E-P72 mice (Fig. 2B, C). Tumors from both E-R72 and E-P72 mice had comparably low levels of apoptosis (Fig. 2D, E). As shown in Fig. 2F, variation in tumor progression was not due to differential expression of Erbb2/Neu and/or p53.

Increased proportion of senescent cells in susceptible mammary glands from E-R72 mice
The local milieu in susceptible tissues can play a critical role in tumor initiation and progression [9,27]. Previous studies have shown that the codon 72 variants differ in their ability to regulate biological processes that regulate cell survival, and modulate the tissue environment [8,[12][13][14][15]27]. Apoptosis, cell cycle arrest and senescence in the mammary glands of adult E-P72 and E-R72 mice were examined to determine the effect of the codon 72 variants on these processes.
First, expression levels of the p53-regulated proapoptotic genes that trigger mitochondrial permeabilization leading to Caspase cleavage were assessed [8]. As shown in Fig. 3A, expression levels of Puma, Noxa, and Bax were similar in mammary glands from E-P72 and E-R72 animals. At the histological level, immunolocalization of CC3 revealed comparably low levels of apoptosis in glands of E-R72 and E-P72 mice (Fig. 3B), indicating that the differences in tumorigenesis were not due to variations in programmed cell death.
Two major effectors of cell cycle arrest and senescence, p21 and p16 INK4a , mediate their effects by disrupting the formation of Cyclin-CDK complexes and activating RB via hypophosphorylation [28][29][30]. The active form of RB binds to members of the E2F family, repressing expression of cell cycle progression genes [31]. As shown in Fig. 3C, mRNA expression levels of p21 and p16 INK4a were two and three times higher, respectively, in the glands of E-R72 compared to E-P72 animals (*p < 0.05). Mammary glands from E-R72 mice also had highly increased p21 protein expression and significantly decreased levels of phosphorylated RB (P-RB) compared to their E-P72 counterparts (Fig. 3D, E).
Sudan Black B (SBB) identifies senescent cells by binding to lipofuscins, which are aggregates of oxidized  proteins and lipids [32]. As shown in Figs. 3F and G, the percentage of SBB+ cells was significantly higher in mammary glands of E-R72 compared to E-P72 animals. These results show that while mammary tumor incidence and proliferation were elevated in E-R72 mice, mammary glands from these animals also had an increased proportion of senescent cells.
Increased SASP, proinflammatory cytokines, and angiogenic factors in susceptible mammary glands of E-R72 mice Traditionally, permanent cell cycle arrest, a pre-requisite of cellular senescence, has been thought to inhibit tumorigenesis by preventing expansion of the population of transformed cells. However, recent studies provide strong evidence to the contrary, demonstrating that senescent cells can acquire a senescence-associated secretory phenotype (SASP), which is proinflammatory and protumorigenic.
SASP is characterized by secretion of a mixture of proinflammatory cytokines, growth and matrix remodeling factors and chemoattractants that contribute to chronic tissue inflammation [28,33]. This inflammatory milieu has been shown to stimulate cancer initiation and progression, due to mutagenic and proliferative signals, respectively [34,35]. RelA (p65), one of five transcription factors in the NFκB signal transduction pathway, is phosphorylated in response to stress and regulates the transcription of factors that contribute to the inflammatory SASP [36,37]. As shown in Figs. 4A and B, levels of phosphorylated p65 (serine 536) were significantly higher in the glands from E-R72 compared to E-P72 mice, demonstrating differential activation of the NFκB pathway. In addition, glands from E-R72 mice had increased expression of genes that regulate acquisition of the secretory phenotype. Expression of Pai1, an angiogenic factor that also contributes to sustained cellular senescence [38], was significantly increased in glands of E-R72 compared to E-P72 mice (Fig. 4C). The matrix remodeling protease gelatinase-B (Mmp9) is associated with breast cancer risk [39], and Mmp9, but not Mmp3, was elevated in glands from R72 mice. Expression of the proinflammatory cytokines Tnfα and Il6, but not Il8, were also significantly increased in mammary glands of E-R72 mice compared to their E-P72 counterparts (Fig. 4C).
Inflammation also stimulates vascular dilation and increased capillary density as part of the immune response [40]. VEGFA is one of the best studied angiogenic factors,  and variants have been associated with increased risk of a variety of cancers, including BrCa [41]. A significant increase in Vegfa expression was observed in the glands of E-R72 mice (Fig. 4C). Vascular density, assessed by immunohistochemical localization of the endothelial cell marker CD31 [42], was also significantly increased (Fig. 4D, E).

Influx of proinflammatory macrophages in susceptible glands of E-R72 mice
The influx and persistence of proinflammatory macrophages are also critical indicators and contributors to chronic tissue inflammation [43]. CCL2 is a major driver of macrophage infiltration and has been shown to promote tumor L. Indirect multiplex immunofluorescence image of macrophages forming a crownlike structure. Hoechst (blue), IBA1 (red), and IL1β (green). Magnification 40x, scale bar 50 µm. Inset: IBA1+ IL1β+ macrophages, white box identifies the magnified area. All values were standardized to the mean of E-P72 samples, and reported as mean ± SEM progression in several cancer models [44]. Expression of this chemoattractant was increased in mammary glands of E-R72 compared to E-P72 animals, by 3.3 times at the message level (Fig. 5A), and 1.5 times at the protein level (Figs. 5B and C).
Macrophage influx was evaluated using the panmacrophage marker, IBA1. Immunohistochemical localization of IBA1 revealed a significantly increased number of macrophages surrounding the ducts (Fig. 5D, E) and within the white adipose tissue (WAT) (Fig. 5D-F) of susceptible mammary glands from E-R72 compared to E-P72 mice. An increase in the number of crown-like structures (CLSs), formed when macrophages surround adipocytes, was also detected in glands from E-R72 animals compared to E-P72 glands (Fig. 5D-G). CLSs are often observed in chronically inflamed breast tissue from obese women, as well as in mammary tissues from obese mice [45,46] and women with breast cancer [47].
Proinflammatory macrophages produce cytokines such as IL1β, and induce reactive nitrogen species (RNS) through upregulation of inducible nitric oxide synthase (iNOS) [48,49]. Expression of Il1β and iNos were significantly upregulated in mammary glands of E-R72 mice (Fig. 5H). Dual immunofluorescent localization of IBA1 and IL1β revealed a significant increase of proinflammatory macrophages around mammary ducts ( Fig. 5I and J) and adipocytes (Fig. 5I-K) in glands of E-R72 compared to E-P72 mice. As shown in Fig. 5L, macrophages in CLSs also expressed high levels of IL1β, providing further evidence of elevated and persistent inflammation in the susceptible glands of E-R72 compared to E-P72 mice.
Increased tumor-associated macrophages (TAMs) and vascular density in mammary tumors from E-R72 mice Recruitment of macrophages to primary tumor sites is essential for tumor progression, and inhibition of CCL2 significantly reduces macrophage infiltration and mammary tumor growth in MMTV-PyMT mice [50]. TAMs promote an immunosuppressive and angiogenic environment that further stimulates tumor growth and progression [51,52]. Immunohistochemical analysis revealed significantly greater numbers of IBA1+ macrophages in mammary tumors from E-R72 compared to E-P72 animals (Figs 6A, B).
TAMs also secrete VEGFA, which promotes angiogenesis and supports tumor progression [49,52]. As shown in Fig. 6C, Vegfa gene expression was elevated in tumors from E-R72 compared to E-P72 mice. Density of intratumoral blood vessels, assessed by the immunolocalization of CD31, was also elevated in mammary tumors from E-R72 animals (Fig. 6D, E). Together, these results show that influx of TAMs, along with elevated levels Vegfa and angiogenesis, contributed to the enhanced tumor progression in E-R72 mice.
Enhanced binding of R72 to promoters of cell cycle arrest and inflammation genes P53 directly transactivates Tnfα, Ccl2, and p21 by binding to canonical p53 response elements (REs) in their respective promoter regions [19,53]. As discussed above, these genes are critical effectors of cellular senescence, SASP and inflammation, and their expression was significantly upregulated in susceptible mammary glands from E-R72 animals (Figs. 3C and 4C). In vivo ChIP-QPCR analysis was used to examine the association of the p53 variants with the REs of their target gene promoters. ChIP assays of extracts from susceptible mammary glands revealed a significantly greater enrichment of the R72 variant at the gene promoters of p21 and Tnfα (Figs. 7A, B, D and E). For Ccl2, binding to RE 2 was similar in both genotypes, however, p53 binding to RE1 was significantly higher in E-R72 animals compared to E-P72 animals (Figs. 7C and F). These results indicate that the R72 variant had an increased affinity for specific REs, resulting in an increased ability to bind to and transactivate these critical genes involved in cellular

Discussion
This study examined the impact of the p53 codon 72 polymorphic variants on mammary tumorigenesis in a humanized ("knock-in") mouse model. Results show that the codon 72 p53 polymorphisms differentially impacted cancer susceptibility and latency in both the DMBAinduced and MMTV-Erbb2/Neu mammary tumor models, representing two distinct tumor etiologies. Compared to the P72 variant, increased mammary tumor incidence and reduced latency were seen in mice homozygous for the R72 variant in both models.
Analysis of susceptible mammary glands revealed an increase in accumulation of senescent cells and SASPmediated chronic inflammation in E-R72 glands compared to their E-P72 counterparts. Cell cycle regulators p16 INK4A and p21, which are effectors of senescence [28,29], were also upregulated in mammary glands of E-R72 mice. R72 showed a greater affinity for the p53 RE in the distal portion of the p21 promoter compared to P72, indicating a direct role of R72 in mediating accumulation of senescent cells in the tissue.
SASP in E-R72 mice was further stimulated by increased expression of Pai1, an effector of senescence [38]. Activation of NFκB, which initiates SASP in senescent cells, was also significantly elevated in mammary glands of E-R72 mice. NFκB pathway activation results in increased expression of proinflammatory cytokines that have been shown to induce a chronically inflamed tissue milieu [36].  Fig. 7 Enhanced binding of R72 to promoters of cell cycle arrest and inflammation genes. Location of p53 response elements (REs) in the distal promoters of A. p21, B. Tnfα, and C. Ccl2 genes, relative to exon 1 (denoted as + 1). Comparisons are of mammary gland extracts from age-matched E-P72 and E-R72 mice. ChIP analysis of p53 binding to REs in the promoters of D. p21, E. Tnfα, and F. Ccl2 genes. All values were standardized to the mean of E-P72 samples, and reported as mean ± SEM (n = 3, *p < 0.05) Proinflammatory cytokines Tnfα and Il6, but not Il8, were upregulated in the chronically inflamed E-R72 glands. Our results also show that R72 played a direct role in increasing expression of Tnfα, since a significantly greater enrichment of the R72 variant of p53 was detected on the Tnfα promoter.
Increased influx of proinflammatory macrophages is also a hallmark of chronic inflammation [54]. In mammary glands of R72 mice, CCL2 was significantly elevated, and there was increased association of the R72 variant with the p53 RE in the distal promoter of the Ccl2 gene. Results also revealed an increased influx of IBA1+ IL1β+ proinflammatory macrophages into the mammary glands of E-R72 mice, which contributed to a chronically inflamed tissue milieu. During chronic inflammation, proinflammatory macrophages produce reactive nitrogen species (RNS) through induction of iNOS, resulting in mutagenic oxidative DNA damage, which contributes to genomic instability [43]. In colorectal cancer, RNS-initiated DNA damage accelerates loss of Apc and enhances tumor development [55]. Our results also show that iNos expression was increased in the susceptible mammary glands of E-R72 mice, indicating that production of RNS enhanced mutagenic transformation of mammary epithelium.
Another contributor to the protumorigenic milieu is angiogenesis, which is essential for maintenance of chronic inflammation [40]. MMP9 increases the bioavailability of VEGFA, which induces angiogenesis and vasodilation [56]. Both Vegfa and Mmp9 were significantly elevated in mammary glands from E-R72 compared to E-P72 mice, along with an increase in CD31+ blood vessels. Increased angiogenesis is seen in susceptible mammary tissue in mouse models that overexpress Cox2 or Ccl2 [57,58], and reduction of angiogenesis by genetic ablation of Cox2 significantly reduces MMTV-Erbb2/Neu-induced mammary tumor incidence [59].
The proinflammatory activities of R72 have also been reported in the Hupki model, particularly when mice are maintained on a high-fat diet. The R72 variant more effectively transactivates Tnfα and Ccl2 genes compared to P72, inducing inflammation in the liver and WAT. This effect is accompanied by an increased proportion of senescent islet cells in the pancreas of R72 compared to P72 mice. In the liver of these Hupki mice, expression of Tnfα and Npc1l1 genes is higher in R72 than P72 mice, where they induce inflammation and increase cholesterol absorption, respectively, predisposing R72 animals to obesity and insulin resistance [19]. These results are in agreement with our findings that the R72 variant more effectively induced inflammation and associated disease progression in the mammary gland.
Interestingly, a recent human epidemiological study demonstrates an association of the R72 variant with inflammatory irritable bowel disease (IBD), which is an important risk factor for colorectal cancer (CRC) [60]. Consistent with this, positive association studies from Iran [61], Greece [62], and Argentina [63] have shown that R72 predicts increased susceptibility to CRC. On the other hand, studies conducted in Turkey [64] and Malaysia [65] have found that P72 is associated with increased risk, indicating a possible role for unique and variable gene-environment interactions, including diet, obesity, and other exposures that affect chronic inflammation.
The association of the p53 codon 72 SNP with BrCa risk is complex. Several studies report that mutated R72 is preferentially retained in breast, colorectal, and head and neck cancers [21,22,66], where it reduces apoptosis, and contributes to therapeutic resistance and tumor progression [67]. The retained, mutated R72 allele is also associated with decreased disease-free interval and overall survival in BrCa [21]. Epidemiological studies have found that the R72 variant is associated with BrCa risk and reduced overall survival [11], and some of the strongest associations are in cohorts of Asian patients [20,24]. Other studies, conducted in predominantly White patient populations, have found that the P72 variant is associated with increased BrCa risk [68]. Taken together, these results indicate that further studies with diverse patient populations are needed to clarify the role of the 72 SNP in BrCa susceptibility globally.
Our studies in a humanized mouse model demonstrated that the p53 codon 72 polymorphic variants differentially affected susceptibility and kinetics of mammary tumorigenesis, and provided evidence for a mechanistic link between R72 and chronic, protumorigenic tissue inflammation. These studies also illustrate that single-nucleotide polymorphisms can have a profound effect on tumorigenesis in vivo. Given the adverse role of chronic inflammation in tumorigenesis, codon 72 variants may predict, at least in part, breast cancer susceptibility, disease progression and/or treatment outcome.

Maintenance and genotyping of mice
Transgenic mice used for these studies were on an FVB background and were homozygous for the proline (P72) or arginine variant (R72). PCR-based genotyping of mice was performed as described previously [17]. Genotypes were verified by a custom quantitative reverse transcription PCR (RT-PCR) SNP allelic discrimination assay (AH89RLW, Thermo Fisher Scientific, Waltham, MA) using an ABI7900HT real-time PCR instrument (Applied Biosystems, Inc., CA). All study mice were generated from genotype-confirmed parents.
Breeders were maintained on standard chow ad libitum, and tumorigenesis study animals were switched to AIN76A (F1515, Bio-Serv, NJ) diet after weaning. Mice were provided free access to drinking water and were housed in Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC)-certified, temperature-and humidity-controlled facilities, with a 12-h light/12-h dark cycle. All procedures were performed according to protocols approved by the Institutional Animal Care and Use Committee at Texas A&M University, College Station, TX. Generation of MMTV-Erbb2/Neu P72R variant mice FVB/N-Tg(MMTVneu)202Mul/J mice were purchased from Jackson Laboratories (Bar Harbor, ME) and bred with p53 polymorphic mice to generate bigenic mice that were hemizygous for Erbb2 and homozygous for the p53 variants. Erbb2 copy number was determined by quantitative RT-PCR according to the Jackson Laboratories protocol (https://www2.jax.org/protocolsdb/f?p = 116:5:0::NO:5: P5_MASTER_PROTOCOL_ID, P5_JRS_CODE:30062,002376). A total of 54 E-P72 and 56 E-R72 mice were used for this study.

Tumorigenesis experiments
Large cohorts of non-randomly assigned animals were used in tumorigenesis studies to ensure adequate statistical power. Animals were monitored daily for health status and tumors in non-blinding manner. Mice were palpated twice weekly throughout the study and sacrificed when the tumor reached 1.5 cm in any direction. Mice were sacrificed by CO 2 asphyxiation, mammary tumors were measured and tissues harvested for later analyses. Tumor histology was evaluated by two independent pathologists.

Chromatin immunoprecipitation (ChIP)
Chromatin from mammary glands of 6-month-old E-P72 and E-R72 mice was isolated as described [75]. In brief, approximately 300 mg of tissue were minced in cold PBS and cross-linked in 1% freshly-made paraformaldehyde-PBS for 10 min. Cross-linking was quenched by adding glycine to a final concentration of 125 mM, and homogenized with a Dounce homogenizer in cold cell lysis buffer (10 mM Tris-Cl, pH 8.0, 10 mM NaCl, 3 mM MgCl 2 , 1% NP-40) supplemented with protease inhibitors (Roche, #04693159001) to generate a single-cell suspension. Cells were incubated on ice, and centrifuged at 1000×g for 10 min at 4°C to pellet nuclei. The pellet was then resuspended in nuclear lysis buffer. Chromatin was sheared by sonication to an average size of 200 to 1000 bp using a Bioruptor (Diagenode, NJ), and insoluble debris was removed by centrifugation. Immunoprecipitation was performed using 10 µg of p53 (FL393) antibody (sc-6243). Primers 5ʹ AAAATCGGAGCTCAGCAGGCCT 3ʹ (forward) and 5ʹ ATCAGGTCTCCACCACCCTGC 3ʹ (reverse) were used for ChIP analysis of the p53 RE in p21 promoter. The quantitative RT-PCR primers used to measure Tnfα and Ccl2 p53 REs were previously published [19].
Sudan Black B (SBB) staining for senescence detection SBB staining was performed on sections of mammary glands from 6-month-old E-P72 and E-R72 animals by the Texas A&M College of Veterinary Medicine histology core laboratory, using the Sudan Black B Histochemical Stain Kit, following the manufacturer's protocol (American MasterTech, KTSBBPT).

Statistical analysis
Log-rank test for trend was performed to statistically compare the Kaplan-Meier curves. Fisher's exact test and Chisquare tests were performed to compare tumor incidence. When animals were sacrificed before the end of study (at 1 year of age), due to causes other than a mammary tumor, they were censored (removed) from the Kaplan-Meier curves on the day of sacrifice. However, the mammary tumor incidence graphs include only animals that had a histologically confirmed mammary tumor, or that survived tumor-free until the end of study. Unpaired two-sided Student's t-test, assuming more conservative unequal variance, was used to compare tumor growth rates, quantitative RT-PCR, IHC, and densitometry analyses. All statistical analyses were performed using GraphPad Prism software (version 6). A p-value < 0.05 was considered statistically significant (*p < 0.05, **p < 0.01 and ***p < 0.001).
Author contributions RTG designed and performed the experiments, interpreted data, and drafted the manuscript. AS contributed to the tumorigenesis studies, including data analysis. LL assisted with generation and interpretation of the confocal images. CN and CJC performed histological assessment of tumors. IL assisted in editing the manuscript, and along with MS was responsible for generating the study animals. RF-Y was responsible for overall conceptual design, study oversight, and editing the manuscript.
Funding This work was funded by the National Institutes of Health grant R01MD006228 (to RF-Y).

Compliance with ethical standards
Conflict of interest The authors declare that they have no conflict of interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/.