Esr1 but Not CYP19A1 Overexpression in Mammary Epithelial Cells during Reproductive Senescence Induces Pregnancy-Like Proliferative Mammary Disease Responsive to Anti-Hormonals

Abstract

Molecular-level analyses of breast carcinogenesis benefit from vivo disease models. Estrogen receptor 1 (Esr1) and cytochrome P450 family 19 subfamily A member 1 (CYP19A1) overexpression targeted to mammary epithelial cells in genetically engineered mouse models induces largely similar rates of proliferative mammary disease in prereproductive senescent mice. Herein, with natural reproductive senescence, Esr1 overexpression compared with CYP19A1 overexpression resulted in significantly higher rates of preneoplasia and cancer. Before reproductive senescence, Esr1, but not CYP19A1, overexpressing mice are tamoxifen resistant. However, during reproductive senescence, Esr1 mice exhibited responsiveness. Both Esr1 and CYP19A1 are responsive to letrozole before and after reproductive senescence. Gene Set Enrichment Analyses of RNA-sequencing data sets showed that higher disease rates in Esr1 mice were accompanied by significantly higher expression of cell proliferation genes, including members of prognostic platforms for women with early-stage hormone receptor–positive disease. Tamoxifen and letrozole exposure induced down-regulation of these genes and resolved differences between the two models. Both Esr1 and CYP19A1 overexpression induced abnormal developmental patterns of pregnancy-like gene expression. This resolved with progression through reproductive senescence in CYP19A1 mice, but was more persistent in Esr1 mice, resolving only with tamoxifen and letrozole exposure. In summary, genetically engineered mouse models of Esr1 and CYP19A1 overexpression revealed a diversion of disease processes resulting from the two distinct molecular pathophysiological mammary gland–targeted intrusions into estrogen signaling during reproductive senescence.

Graphical abstract

Breast cancer continues to be a world-wide challenge.¹ In 2020, it was the most diagnosed malignancy in women. In spite of all the current treatment options, it continues to lead to significant morbidity and mortality. Disparities between populations for access to optimal medical care remain a persistent problem across the globe. Understanding the pathophysiology of breast cancer and dissection of disease pathways could lead to more effective prevention, screening, and early detection strategies.

Unrestrained cell proliferation is an established risk factor for cancer generation, including breast cancer.² Estrogen signaling pathways play a direct role in breast epithelial cell growth-inducing carcinogenesis as well as influencing the surrounding stromal and immunologic environment.^3,4 The increased levels of estrogen receptor α (ER) and aromatase expression found in post-menopausal breast tissue are posited to contribute to the increased risk of breast cancer in post-menopausal women.^5,6 ER contributes to cell proliferation of mammary epithelial cells via both genomic and nongenomic actions.⁷ ER genomic action activates signaling pathways linked to regulation of cell proliferation, including myc proto-oncogene basic helix-loop-helix transcription factor (MYC), cyclin D1, the cyclin E (CCNE1)–cyclin-dependent kinase 2–cyclin-dependent kinase inhibitor 1a complex, and cell survival, including bcl2 apoptosis regulator and bcl2-like 1.^8,9 Proteins implicated in nongenomic extranuclear pathway ER-mediated cell proliferation pathways include insulin-like growth factor I receptor, matrix metallopeptidase 2, matrix metallopeptidase 9, epidermal growth factor receptor (EGFR), mitogen-activated protein kinase, src homology 2 domain-containing adaptor protein 1, growth factor receptor-bound protein 2, and son of sevenless Ras/Rac guanine nucleotide exchange factor 1.^10,11 Aromatase, encoded by the cytochrome P450 family 19 subfamily A member 1 (CYP19A1) gene in humans, acts as the rate-limiting enzyme for conversion of testosterone to estrogen.¹² Breast-localized aromatase expression has been linked to increased breast cancer risk through ER-dependent genomic and nongenomic mechanisms as well as ER-independent actions that include direct DNA damage.^11,12

Anti-hormonal agents with positive results for primary breast cancer prevention in high-risk women include both selective ER modulators, such as tamoxifen and raloxifene, as well as aromatase inhibitors.13, 14, 15 Selective ER modulator–associated adverse effects are one reason research continues into assessment of aromatase inhibitors as well as other agents and preventive interventions.^16,17 Pathways that may modify response to anti-hormonal agents, such as interferon-related pathways, are also a research focus.18, 19, 20, 21

Studies in breast cancer pathogenesis may determine a molecular prognostic risk classification system for primary cancer occurrence, or secondary cancer recurrence, of different types of breast cancers.^22,23 In particular, the Prediction Analysis of Microarray 50 (PAM50) risk of recurrence score has been useful for evaluating prognosis of early-stage hormone receptor–positive/human epidermal growth factor receptor 2-negative breast cancer in women, including in post-menopausal women.^24,25 Development of such a risk classifier may need to incorporate multiple parameters, including transcriptome information.26, 27, 28 For example, the PAM50 panel includes genes linked to genomic (MYC and CCNE1) and nongenomic (EGFR) ER-mediated mechanisms of cell proliferation from cell culture studies. Parallels in transcription between human and mouse models have been used to study disease pathogenesis in the natural setting.^29,30

Breast cancer–related mouse models enable investigation of disease pathophysiology within the context of physiological endocrine and immunologic function.31, 32, 33 In a mouse model of reproductive aging, both mouse mammary tumor virus–reverse tetracycline–controlled transactivator/tet-operator (tet-op)–estrogen receptor 1 (Esr1) and mouse mammary tumor virus–reverse tetracycline–controlled transactivator/CYP19A1 (CYP19A1) mice show elevated expression levels of Esr1 and Pgr.^19,34 Preneoplastic lesions in both models are typically estrogen receptor α positive (ER⁺) in prereproductive senescent mice, but invasive cancers are either negative (ER⁻) or have a low percentage of ER positivity in the cancers.34, 35, 36 Prereproductive senescent Esr1, but not CYP19A1, mice are relatively resistant to tamoxifen,³⁵ which is linked to increased activation of an interferon regulatory factor 7 (IRF7)-STAT1 pathway in mammary epithelial cells.¹⁹

When performing transcriptional studies in whole mammary tissue, expression levels of luminal [keratin (Krt) 7, Krt8, Krt18, and Krt19) and basal (Krt5 and Krt14) cytokeratins can be used for assessment of relative mammary epithelial cell content in samples.37, 38, 39 Levels have to be understood in context, as keratins such as Krt5 and Krt14 can both contribute to regulation of cell proliferation and exhibit higher expression levels in proliferative cells.^37,40 Expression levels of Krt5, Krt14, and Krt17 are included in the PAM50 risk panel.²⁵

In mice, the major source of estrogen production is the granulosa cells of the ovarian follicles.⁴¹ Reproductive senescence is initiated by loss of ovarian follicles with coincident decrease in systemic estrogen levels. Ovarian follicle development, including numbers and types of follicles, is assessed histologically in ovarian tissue samples as a measure of ovarian follicle reserve and reproductive function.^42,43 When ovarian follicle numbers are low, systemic estrogen levels are also low.⁴⁴

The current study was initiated to determine whether altering the induction of Esr1 and CYP19A1 transgene expression to after reproductive maturity, and then postponing anti-hormonal therapy until reproductive senescence, could differentially impact disease pathophysiology and anti-hormonal responsiveness in the two models. Aging is the greatest risk factor for developing breast cancer.45, 46, 47 Studies indicate that >80% of cases occur in women aged >50 years, 50% in women aged >65 years, and 21% in women aged >70 years.48, 49, 50, 51 Several mechanisms have been suggested to account for this relationship, including alterations in progenitor cell populations, age-associated somatic mutation, changes in the microenvironment, and epigenetic alterations.⁵¹ Yet, despite the association of aging with breast cancer, most mouse model studies of breast cancer genetics are conducted before reproductive senescence.^32,52, 53, 54 Notable exceptions include mice with cyclin D1, nuclear receptor coactivator 3 (alias amplified in breast cancer 1 protein; AIB1), or prolactin overexpression or Stat1 deletion, where most invasive cancers arise with or after reproductive senescence.55, 56, 57, 58 A unique aspect of this study is the use of conditional transgene models. Genetic alterations in most nonconditional models are present from birth and active during early during reproductive life. In this study, the Esr1 and CYP19A1 transgenes were induced at middle age and continued through reproductive senescence as the mice aged. The impact of two different anti-hormonals, tamoxifen and letrozole, both prescribed for breast cancer risk reduction in post-menopausal women, was studied in the mice during reproductive senescence. Significantly, the study revealed a differential impact of Esr1 overexpression versus CYP19A1 overexpression at this age. Compared with CYP19A1 mice, Esr1 mice exhibited a significantly higher proliferative transcriptional response, reminiscent of pregnancy, with parallels to the human PAM50 risk profile that associated with increased prevalence of preneoplasia and cancer. Esr1-induced proliferative changes resolved to levels approximating those found in the less-proliferative CYP19A1 mice following tamoxifen or letrozole. The study demonstrated that increased breast cancer risk due to ER or aromatase overexpression can be effectively modeled, and interventions can be tested in aging mice.

Materials and Methods

Mouse Models

The Georgetown University (Washington, DC) Institutional Animal Care and Use Committee approved the animal research protocol, and all regulations concerning the use of animals in research were carefully adhered to throughout the conduct of the experiments. Mouse mammary tumor virus–reverse tetracycline–controlled transactivator/tet-op–Esr1 and mouse mammary tumor virus–reverse tetracycline–controlled transactivator/tet-op-CYP19A1 mice on a C57Bl/6 background were bred in the Georgetown University Department of Comparative Medicine facility, genotyped at weaning (Transnetyx, Inc., Cordova, TN), and sequentially placed through the four different experimental cohorts for each genotype until each cohort was filled with end point cohort size planned for n = 20. Twelve C57Bl/6 wild-type mice were included as a reference cohort for ovarian follicle counts between 2 and 5 months of age (n = 2 age 2 months, n = 3 age 3 months, n = 5 age 4 months, and n = 2 age 5 months). A total of 10% and 25% losses from the cohorts were predicted for the 18- and 20-month end point cohorts, respectively (https://www.nia.nih.gov/research/dab/aged-rodent-colonies-handbook/animal-information, last accessed September 17, 2022). All cohorts were raised on Department of Comparative Medicine standard laboratory mouse chow until 12 months of age, when they were switched to diet containing 200 mg doxycycline per kilogram food (Bio-Serv, Flemington, NJ) for transgene induction. One cohort from each genotype was euthanized for the age 18-month end point. The three 20-month end point cohorts from each genotype were subjected to anesthesia and surgical s.c. pocket formation with placement of no pellet, a tamoxifen pellet (25 mg/60-day release), or a letrozole pellet (2.5 mg/60-day release; Innovative Research of America, Sarasota, FL) at age 18 months and followed up until age 20 months, when they were euthanized. Euthanasia was conducted according to the approved animal protocol using CO₂ inhalation followed by cervical dislocation. All mice were individually followed up for development of disease or cage death as they were aged to the 18 and 20 months of age end points. Mice were euthanized before the age 18- or 20-month experimental end points according to the approved research protocol for pain/distress/infection unresponsive to treatment, severe lethargy or weakness, severe neurologic signs, severe respiratory distress, total tumor burden >2 cm, weight loss ≥15% of body weight, or wounds refractory to treatment. Cage deaths were unpredicted spontaneous deaths lacking an identifiable cause. Numbers of mice in each cohort were as follows: Esr1: 18 month n = 26 entered, n = 3 early necropsy, n = 3 cage death, end point n = 20 (77% survival); 20 month n = 33 entered, n = 9 early necropsy, n = 5 cage death, end point n = 19 (58% survival); 20 month with tamoxifen n = 32 entered, n = 4 early necropsy, n = 4 cage death, end point n = 24 (75% survival); 20 month with letrozole n = 32 entered, n = 11 early necropsy, n = 1 cage death, n = 20 (62% survival). CYP19A1: 18 month n = 22 entered, end point n = 22 (100% survival); 20 month n = 32 entered, n = 5 early necropsy, n = 9 cage death, end point n = 18 (56% survival); 20 month with tamoxifen n = 31 entered, n = 5 early necropsy, n = 3 cage death, end point n = 23 (74% survival); 20 month with letrozole n = 33 entered, n = 2 early necropsy, n = 7 cage death, n = 24 (73% survival). Mean weights ± SEM at end point for each cohort were as follows: Esr1: 18 month 41 ± 3 g, 20 month 50 ± 3 g, 20 month with tamoxifen 45 ± 3 g, 20 month with letrozole 46 ± 4 g. CYP19A1: 18 month 32 ± 1 g, 20 month 31 ± 1 g, 20 month with tamoxifen 31 ± 1 g, 20 month with letrozole 33 ± 1 g.

To validate transgene expression in the different cohorts, thoracic (number 2) mammary glands were flash frozen in liquid nitrogen and stored at −80°C until processing. Randomly selected glands from each cohort were thawed in Invitrogen Trizol reagent (ThermoFisher Scientific, Waltham, MA) and homogenized using a Qiagen Tissuruptor and Qiashredder (Qiagen, Hilden, Germany). RNA was then isolated using Direct Zol RNA miniprep kit (Zymo Research, Irvine, CA), quantified on a Qubit 2.0 Fluorometer (Qiagen) and analyzed for RNA integrity using the Agilent 2100 Bioanalyzer (Agilent, Santa Clara, CA). RNA (1 μg) was used to generate cDNA with the High Capacity RNA to cDNA kit (Applied Biosystems, Waltham, MA), and RT-PCR was conducted using Platinum Taq DNA Polymerase (Invitrogen, Waltham, MA). For tet-op–Esr1, two unique primer pairs were used and expression was normalized to β-actin expression levels⁵⁹: Sp3 pair: 5′-CCACACCAGCCACCACCTTC-3′ (forward) and 5′-CCACTTCAGCACATTCCTTA-3′ (reverse); and Sp4 pair: 5′-GATGAGACAGCACAACAACC-3′ (forward) and 5′-CAAAGGCATGGAGCATCTCT-3′ (reverse); predicted sizes were 287 and 385 bp, respectively. For tet-op–CYP19A1, one unique primer set was used and expression was normalized to β-actin expression levels³⁴: tet-op–CYP19A1: 5′-CCTTGCACCCAGATGAGACT-3′ (forward) and 5′-GACAGCACAACAACCAGCAC-3′ (reverse); predicted size was 134 bp. Endogenous mouse β-actin was 5′-ATCGTGGGCCGCCCTAGGCA-3′ (forward) and 5′-TGGCCTTAGGGTTCAGAGGG-3′ (reverse); predicted size was 244 bp. PCR amplicons were run on 2% agarose E-gels (Invitrogen) and imaged using blue fluorescence on the Amersham Imager 600 (GE HealthCare, Chicago, IL). Adobe Photoshop (San Jose, CA) was used to calculate relative pixel intensities for the bands from each image, with normalization of transgene expression levels to β-actin expression level individually for each sample. Percentage relative transgene expression was calculated across all four cohorts for each genotype, setting the age 18-month expression levels at 100%.

Mammary Gland Whole Mount Evaluation

Carmine-alum–stained mammary gland whole mounts were prepared from one inguinal (number 4) mammary gland from each necropsied mouse at end point using standard procedures.⁶⁰ Whole mounts were visually examined, and images were taken at ×0.5 utilizing a Nikon Eclipse E800M microscope with Nikon DMX1200 software (Nikon Instruments, Inc., Melville, NY). Four independent observers (P.A.F., W.W., B.L.R., and V.M.) blindly scored whole mounts for branching structures (secondary versus tertiary),⁶¹ lobular growth (presence versus absence),⁶² and hyperplastic alveolar nodules (HANs; number present).⁶³ The final score was the majority score. In the case of a tie, images were re-examined by P.A.F. and a final score was set. Glands were defined as hyperplastic when at least one HAN was present in the whole mount. Presence or absence of dense uniform pregnancy-like growth patterns was scored by P.A.F. Mammary gland images were digitally scored for mean relative density utilizing an automated program that uses mean pixel intensity to represent mammary density.⁶⁴ Lower-density scores correspond to lower mean pixel intensity scores and higher mammary gland density. Higher-density scores correspond to higher mean pixel intensity and lower mammary gland density.

Histology and ER α Immunohistochemistry

Inguinal (number 4) mammary glands were fixed in 10% neutral formalin solution and embedded in paraffin, and tissue sections (5 μm thick) were utilized for hematoxylin and eosin (H&E) staining and ER α immunohistochemistry. Relative cellularity was assessed on H&E-stained mammary mid-gland longitudinal sections. Low cellularity was defined as ≤10 cell clusters, moderate cellularity was defined as 11 to 20 clusters, and high cellularity was defined as >20 clusters. A cell cluster was defined as five or more cells. ER α protein was detected utilizing rabbit polyclonal anti–estrogen receptor α (catalog number 06-935; Millipore Sigma, Burlington, MA), 1:4800, 1 hour, room temperature. Before application of primary antibody, heat-induced epitope retrieval was performed by immersing tissue sections at 98°C for 20 minutes in 10 mmol/L citrate buffer (pH 6.0) with 0.05% Tween, followed by treatment with 3% hydrogen peroxide and 10% normal goat serum (10 minutes each). Following primary antibody exposure, slides were exposed to a horseradish peroxidase–labeled polymer [30 minutes; EnVision+ System-HRP Labeled Polymer Anti-Rabbit; K4003 (Agilent, Dako, Carpinteria, CA)] and 3′-diaminobenzidine (5 minutes; Agilent, Dako) and counterstained with hematoxylin (Harris Modified Hematoxylin; Fisher Scientific, Hampton, NH). H&E slides were read blindly by two independent readers (P.A.F. and W.W.). ER α immunohistochemistry was read by P.A.F. Digital images were taken at ×40 utilizing a Nikon DMX1200 camera mounted on a Nikon Eclipse E800M microscope.

RNA-Sequencing Analysis and Visualizations

Thoracic (number 2) mammary glands were flash frozen in liquid nitrogen and stored at −80°C until processing. Glands were thawed in Invitrogen Trizol reagent (ThermoFisher Scientific) and homogenized using a Qiagen Tissuruptor and Qiashredder. RNA was then isolated using Direct Zol RNA miniprep kit (Zymo Research), quantified on a Qubit 2.0 or 4.0 Fluorometer (Qiagen) and analyzed for RNA integrity using the Agilent 2100 Bioanalyzer. Indexed, single-index sequencing libraries were prepared from 1 μg ribosome depleted total RNA using TruSeq Stranded Total RNA Library Preparation Human/Mouse/Rat (Illumina, San Diego, CA). Sequencing was performed using the Illumina NextSeq 550, SE 75-bp read length; minimum reads ≥50 million per sample. Sequencing quality was checked using FastQC processing version 0.11.9 (https://www.bioinformatics.babraham.ac.uk/projects/fastqc, last accessed August 7, 2022). Contaminated adaptor and/or low-quality portions of sequenced reads were trimmed using Trim Galore version 0.6.5 (https://github.com/FelixKrueger/TrimGalore, last accessed August 7, 2022). Trimmed reads were aligned to the reference mouse genome (mm10), using STAR version 2.7.9a.⁶⁵ Batch effect was normalized using RUVSeq version 1.26.0 with the RUVg method.⁶⁶ Normalized expression levels were estimated by means of transcripts per million using RSEM version 1.3.1.⁶⁷ Differentially expressed genes (DEGs) between genotypes at each end point were identified using DESeq2.⁶⁸ Genes were considered statistically significantly differentially expressed when the adjusted P value was <0.05. Protein-coding DEGs between models at each end point were identified, and numbers were visualized as expressed higher in Esr1 versus CYP19A1 mice using bar graphs with each group analyzed independently for enrichment in HALLMARK gene sets (Gene Set Enrichment Analysis, Molecular Signatures Database version 7.5.1, updated January 2022, last accessed July 31, 2022).^69,70 The top three gene sets by false discovery rate q-value were visualized using bar graphs. Because the highest numbers of DEGs between the two models occurred at the 20-month end point with identification of cell proliferation–related enriched HALLMARK gene sets in the Esr1 mice, DEGs from this group were selected to identify overlaps with the Prosigna PAM50 gene set.²² Transcripts per million for overlapping genes were presented by bar graph, categorized by cell proliferation versus estrogen response–related genes. A heat map illustrating changes in relative gene expression for these genes across all four end points for both models was constructed (GraphPad Prism version 9.3.1; GraphPad Software, San Diego, CA). A total of nine DEGs at the 20-month end point identified as members of both the PAM50 profile are known to exhibit pregnancy stage–related changes in gene expression.¹¹ Heat maps of these nine as well as 34 additional genes regulated during normal pregnancy-related development from National Center for Biotechnology Information's Gene Expression Omnibus GSE70440 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE70440, last accessed September 17, 2022)⁷¹ were constructed to visualize relative expression levels during normal pregnancy²⁸ in comparison to relative expression levels across time and treatment condition for the eight experimental cohorts (GraphPad Prism version 9.3.1). A heat map of 12 Irf7-Stat1 immune-related genes linked to tamoxifen resistance,¹⁹ five HALLMARK_INFG_RESPONSE, and four HALLMARK_TNFA_SIGNALING_VIA_NFKB DEGs identified at age 12 months in Esr1 and CYP19A1 mice with transgene induction at age 12 months was constructed to illustrate the differences in relative levels of gene expression between the tamoxifen-resistant 12-month–old Esr1 mice and tamoxifen-responsive 18- and 20-month–old Esr1 and 12-, 18-, and 20-month–old CYP19A1 mice (GraphPad Prism version 9.3.1). The new data discussed in this publication were deposited in National Center for Biotechnology Information's Gene Expression Omnibus⁷² and are accessible through series accession number GSE201326 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=gse201326, last accessed August 7, 2022).

Ovarian Follicle Counts

Ovaries were subjected to histologic evaluation of healthy and unhealthy follicles.^42,43 The ovaries were fixed in paraformaldehyde, embedded in paraffin, and sectioned every 8 μm using a microtome. After sectioning, slides were stained with H&E. Every 10th section of the ovary was used to count the total number of primordial follicles, primary follicles, preantral follicles, antral follicles, atretic follicles, and abnormal follicles. All sections were examined without knowledge of treatment group. Primordial follicles were defined as follicles with an oocyte, surrounded by a single layer of squamous granulosa cells. Primary follicles were defined as follicles that consist of an oocyte, surrounded by a single layer of cuboidal granulosa cells. Preantral follicles were defined as follicles containing an oocyte, surrounded by multiple layers of cuboidal granulosa cells and theca cells. Antral follicles were defined as follicles that consist of an oocyte, surrounded by numerous layers of cuboidal granulosa cells, theca cells, and a fluid-filled antrum. Atretic follicles were counted as preantral or antral follicles that contain ≥10% number of apoptotic bodies. Abnormal follicles included follicles with double oocytes and/or fragmented nuclei. Preantral and antral follicles were required to have nuclear material present to avoid double counting. Total number of follicles and number and percentage of each type of follicle were recorded.

Statistical Analysis

Calculations of means and SEMs were performed utilizing GraphPad Prism version 9.3.1. Multiple unpaired t-tests were used to compare primordial, primary, preantral, antral, and atretic follicle counts; the Fisher exact test was used to determine if there were nonrandom associations between tertiary and secondary branching, presence and absence of lobular growth, HANs, or preneoplasia/cancer; and U-test was used to compare relative mean mammary gland density scores (P < 0.05 considered statistically significant; GraphPad Prism version 9.3.1). The Fisher exact, Freeman-Halton extension was used to determine if there were nonrandom associations between proportions of mice with cancer, preneoplasia, and normal findings between cohorts (P < 0.05 considered statistically significant; http://vassarstats.net/fisher2x3.html, last accessed August 7, 2022). Scatterplots and bar graphs were prepared utilizing GraphPad Prism version 9.3.1. Significance levels for specific comparisons are indicated by asterisks: ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, and ∗∗∗∗P < 0.0001.

Results

Changes in Mammary Gland Morphology Induced by Reproductive Senescence in Mouse Models of Breast Cancer Risk Are Accentuated by Tamoxifen and Letrozole Exposure

The study design was intended to model how increased levels of Esr1 and CYP19A1 expression during menopause in women impact response to anti-hormonals, such as tamoxifen and letrozole, prescribed for breast cancer risk reduction. Esr1 and CYP19A1 transgenes targeted to mammary epithelial cells were induced at age 12 months, a time point corresponding to middle age in humans (https://www.jax.org/research-and-faculty/research-labs/the-harrison-lab/gerontology/life-span-as-a-biomarker, last accessed August 7, 2022).⁷³ At weaning, mice were divided into four cohorts: one with an age 18-month end point and three cohorts with surgery at age 18 months for placement for tamoxifen or letrozole or no pellet placement and then followed for 2 months with end point at age 20 months (Figure 1A). Ovarian follicle counts were performed on mice from each cohort to assess ovarian reserves as a marker of reproductive senescence. All eight experimental cohorts demonstrated significantly fewer counts of all follicle types which were abnormal compared with young reproductive age wild-type mice (age 2 to 5 months), indicating that both the 18- and 20-month–old mice had entered into reproductive senescence (Figure 1B). Mammary gland whole mounts were used to assess branching, lobular growth morphology, and overall mammary gland density. Changes in branching structure were the least sensitive to reproductive senescence, with only Esr1 mice exhibiting a significant decrease in the proportion of mice with tertiary branching between 18 and 20 months of age (Figure 2A). Significant reductions in lobular growth, accentuated in mice treated with either anti-hormonal, were found with progression through reproductive senescence (Figure 2B). A significant difference between the models was the generally higher prevalence of lobular growth in the Esr1 mice across different ages and treatment groups, with reductions in prevalence with age and anti-hormonal exposure (Figure 2, C–J). Transgene expression was retained in all cohorts despite reductions in epithelial cell content with age, and anti-hormonal treatment was validated by RT-PCR (Supplemental Figure S1A).

Figure 1.

Experimental design and characterization of ovarian follicle counts during reproductive senescence at 18 and 20 months (m) of age. A: Study cohort design: Esr1 and CYP19A1 transgene expression was induced at age 12 months. The end point at age 18 months had no intervention. The end point at age 20 months was with s.c. pellet placement surgery at age 18 months for anti-hormonal exposure [none, tamoxifen (25 mg/60-day release), or letrozole (2.5 mg/60-day release)]. B: Scatterplots illustrate distribution of ovarian follicle counts of Esr1 and CYP19A1 mice at 18 and 20 months of age without and with tamoxifen or letrozole exposure. Follicle counts of wild-type mice at 2 to 5 months of age shown for reference. Black fill: Esr1. White fill: CYP19A1. Circles: age 18m (n = 18 Esr1, n = 22 CYP19A1). Squares: age 20m (n = 11 Esr1, N = 10 CYP19A1). Inverted triangle: age 20m with 2 months tamoxifen exposure (n = 9 Esr1, n = 10 CYP19A1). Triangle: age 20m with 2 months letrozole exposure (n = 9 Esr1, n = 11 CYP19A1). Gray circles: wild type (n = 12). Primordial, primary, preantral, antral, and atretic follicle counts significantly lower in all cohorts of 18- and 20-month–old mice compared with 2- to 5-month–old mice (P < 0.005, multiple unpaired t-tests, GraphPad Prism version 9.3.1). Data are presented as means ± SEM (B). MMTV, mouse mammary tumor virus; rtTA, reverse tetracycline–controlled transactivator; tet-op, tet-operator.

Figure 2.

Impact of reproductive senescence and anti-hormonal exposure on patterns of mammary gland branching and lobular growth in mouse estrogen receptor 1 (Esr1) and human cytochrome P450 family 19 subfamily A member 1 (aromatase; CYP19A1) mice. A: Bar graphs illustrating percentage of Esr1 and CYP19A1 mice with secondary and tertiary branching in each cohort. Esr1: age 18 months (m) n = 20; age 20m n = 19; age 20m tamoxifen n = 24; age 20m letrozole n = 20. CYP19A1: age 18m n = 21; age 20m n = 18; age 20m tamoxifen n = 23; age 20m letrozole n = 24. B: Bar graphs illustrating percentage of Esr1 and CYP19A1 mice with and without lobular growth in each cohort. Esr1: age 18m n = 20; age 20m n = 19; age 20m tamoxifen n = 24; age 20m letrozole n = 20. CYP19A1: age 18m n = 21; age 20m n = 18; age 20m tamoxifen n = 23; age 20m letrozole n = 24. C–J: Representative mammary gland whole mounts of 18-month–old Esr1 (C) and CYP19A1 (D) mice, 20-month–old Esr1 (E) and CYP19A1 (F) mice, tamoxifen-exposed 20-month–old Esr1 (G) and CYP19A1 (H) mice, and letrozole-exposed Esr1 (I) and CYP19A1 (J) mice, with right-hand arrow indicating generally decreasing lobular growth prevalence with age and anti-hormonal exposure. All images scaled identically. White asterisks indicate presence of lobular growth to varying degrees in different cohorts. ∗P < 0.05, ∗∗P < 0.01, and ∗∗∗P < 0.001 (Fisher exact test, two sided, GraphPad Prism version 9.3.1). Scale bar = 1000 μm (C–J). Original magnification, ×0.5 (C–J).

Esr1 Mice Show Significantly More Abnormal Mammary Pathology than CYP19A1 Mice during Reproductive Senescence

H&E-stained slides of mammary gland tissue were used to assess cellularity and prevalence of preneoplasia (ductal hyperplasia, lobular hyperplasia, ductal carcinoma in situ, and adenosis) and invasive cancer (adenosquamous carcinoma and adenocarcinoma) during reproductive senescence and with anti-hormonal exposure. Mammary gland whole mounts were used to score the presence or absence of HANs. Relative cellularity decreased with tamoxifen and letrozole in both models but remained significantly higher in Esr1 compared with CYP19A1 mice with tamoxifen exposure (Figure 3A). The percentage of mice with HANs was higher in 18-month–old Esr1 compared with that in CYP19A1 mice (Figure 3B). At age 20 months, Esr1 mice demonstrated significantly more preneoplasia and cancer than CYP19A1 mice, which was abrogated by both tamoxifen and letrozole exposure (Figure 3, C and D). ER immunohistochemistry was used to assess patterns of nuclear-localized ER staining in normal, preneoplastic, and cancerous mammary tissue. Patterns of ER staining in normal-appearing ducts were not significantly altered by age or anti-hormonal exposure (Figure 4, A–C, and Supplemental Figure S1, B–F). Ductal and lobular hyperplasias showed higher percentages of ER⁺ cells compared with normal ducts (Figure 4, D–G) without significant differences between the models. Ductal carcinoma in situ lesions were most commonly ER⁺ (Figure 4, H and I), but an ER⁻ ductal carcinoma in situ lesion was identified in a CYP19A1 mouse (Figure 4J). Adenosis and adenosquamous lesions were ER⁺ in both models (Figure 4, K–P). One ER⁻ adenocarcinoma was found in an Esr1 mouse (Figure 4Q). The other adenocarcinomas were ER⁺ (Figure 4R). In summary, Esr1 mice showed higher prevalence and more advanced mammary pathology compared with CYP19A1 mice; however, the disease itself was histologically similar between models. An important finding was the relatively high prevalence of ER⁺ disease in the Esr1 and CYP19A1 mice during reproductive senescence with initiation of transgene expression at age 12 months that could be reduced by exposure to tamoxifen or letrozole.

Figure 3.

Effect of reproductive senescence and anti-hormonal exposure on mammary gland cellularity, preneoplasia, and cancer in Esr1 and CYP19A1 mice. A: Bar graphs illustrate percentage of mice with low, medium, or high epithelial cellularity in Esr1 and CYP19A1 cohorts. Esr1: age 18 months (m) n = 20; age 20m n = 19; age 20m tamoxifen n = 24; age 20m letrozole n = 20. CYP19A1: age 18m n = 22; age 20m n = 18; age 20m tamoxifen n = 23; age 20m letrozole n = 24. P values determined by Fisher exact test, Freeman-Halton extension (http://vassarstats.net/fisher2x3.html, last accessed August 7, 2022). B: Bar graphs illustrate percentage of mice with at least one hyperplastic alveolar nodule (HAN) detected in Esr1 and CYP19A1 cohorts. Esr1: age 18m n = 20; age 20m n = 19; age 20m tamoxifen n = 24; age 20m letrozole n = 20. CYP19A1: age 18m n = 21; age 20m n = 18; age 20m tamoxifen n = 23; age 20m letrozole n = 24. P values determined by Fisher exact test, two sided, GraphPad Prism version 9.3.1. C: Bar graphs illustrate percentage of Esr1 and CYP19A1 mice with completely normal versus preneoplastic versus cancer findings. Esr1: age 18m n = 20; age 20m n = 19; age 20m tamoxifen n = 24; age 20m letrozole n = 20. CYP19A1: age 18m n = 22; age 20m n = 18; age 20m tamoxifen n = 23; age 20m letrozole n = 24. P values determined by Fisher exact test, Freeman-Halton extension (http://vassarstats.net/fisher2x3.html, last accessed August 7, 2022). D: Bar graphs illustrating numbers of combinations of preneoplasia and cancer found in Esr1 and CYP19A1 mice. Total number of mice in each cohort listed on x axis. P values determined by Fisher exact test, two sided, GraphPad Prism version 9.3.1, number of mice with preneoplasia/cancer versus those with only normal findings. ∗P < 0.05, ∗∗P < 0.01, and ∗∗∗∗P < 0.0001. DCIS, ductal carcinoma in situ; DH, ductal hyperplasia; LH, lobular hyperplasia.

Figure 4.

Comparison of histology and estrogen receptor α (ER) immunohistochemistry in mouse estrogen receptor 1 (Esr1) and human cytochrome P450 family 19 subfamily A member 1 (aromatase; CYP19A1) mice. Representative images of ER immunohistochemistry of normal ducts (A–C), ductal hyperplasia (DH; D and E), lobular hyperplasia (LH; F and G), ductal carcinoma in situ (DCIS; H–J), adenosis (K–M), adenosquamous cancers (N–P), and adenocarcinomas (Q and R). For comparison, representative images of ER immunohistochemistry of normal ducts from CYP19A1 20-month–old, tamoxifen-exposed CYP19A1 20-month–old, and letrozole-exposed Esr1 mice shown in Supplemental Figure S1. Black arrows indicate representative cells with nuclear-localized ER staining. Scale bar = 10 μm (A–R).

Expression of Cell Proliferation Genes Is Significantly Higher in Esr1 Mice during Reproductive Senescence but Is Down-Regulated with Anti-Hormonal Exposure

RNA-sequencing analyses were used to explore transcriptional changes induced by Esr1 compared with CYP19A1 expression during reproductive senescence and with tamoxifen and letrozole exposure. Significantly differentially expressed protein-coding genes (DEGs) were compared for each end point/treatment condition between the two models and subjected to Gene Set Enrichment Analyses. Numbers of DEGs and top three gene enrichment sets for each comparison were determined (Figure 5, A–L). The largest numbers of DEGs between the models were found at age 20 months in the absence of anti-hormonal exposure, with significant enrichment in HALLMARK_G2M_CHECKPOINT, HALLMARK_E2F_TARGETS, and HALLMARK_MITOTIC_SPINDLE in the DEGs expressed at significantly higher levels in the Esr1 mice (Figure 5, D–F). Individual significantly differentially expressed genes identified for each enrichment group with fold change in expression levels between genotypes and individual level of statistical significance are presented in Supplemental Table S1. To explore the relevancy of the pathologic and molecular differences observed in the Esr1 and CYP19A1 models with human breast cancer, identified DEGs were compared with the human PAM50 gene panel to determine whether any of the PAM50 panel genes were differentially expressed between Esr1 and CYP19A1 mice (Figure 6). The specific intent was to determine whether there was a correlation between the higher rate of abnormal pathology in the Esr1 mice with higher expression of any PAM50 genes. Expression levels of genes from the PAM50 profile were largely comparable between models at age 18 months (Figure 6A). However, at age 20 months, 18 cell proliferation–related genes were expressed at significantly higher levels in Esr1 compared with CYP19A1 mice (Figure 6B). Six estrogen response–related genes from the PAM50 gene panel were identified as significantly higher expressed in the CYP19A1 mice compared with the Esr1 mice at age 20 months (Figure 6B). Expression levels following tamoxifen (Figure 6C) and letrozole (Figure 6D) exposure were lower and largely comparable between genotypes. Statistical analyses of relative expression levels within genotypes across age and exposure groups show how exposure to tamoxifen and letrozole was associated with significant down-regulation of cell proliferation genes in Esr1 mice (Supplemental Figure S2), with Foxa1, Pgr, and Sfrp1 being significantly down-regulated in the CYP19A1 mice (Supplemental Figure S3). Relative expression levels of luminal (Krt7, Krt8, Krt18, and Krt19) and basal (Krt5 and Krt14) cytokeratins were compared across genotypes and exposure groups for a general assessment of mammary epithelial cell content and relative populations of luminal and basal mammary epithelial cells in the different samples (Supplemental Figure S4). Luminal cytokeratins were expressed at the same or higher levels in samples from CYP19A1 mice compared with samples from Esr1 mice of the same age and/or exposure group, suggesting mammary epithelial cell content in the samples used for RNA sequencing was reasonably comparable between samples. Significantly higher levels of Krt5 and Krt14, found in Esr1 mice, are consistent with possible differences in mammary epithelial cell populations as well as being compatible with the higher proliferation rates at age 20 months in these mice. Significantly lower luminal Krt expression levels following letrozole were found in both Esr1 and CYP19A1 mice. The analysis of the PAM50 panel highlighted the impact of Esr1 and CYP19A1 overexpression on cell proliferation and estrogen response genes. Both models develop ER⁺ pathology and express Esr1 (Figures 4 and 6). To explore expression of previously reported genes mediating genomic versus nongenomic mechanisms of estrogen receptor–mediated cell proliferation,7, 8, 9, 10, 11 expression patterns of well-established seven genes linked to genomic signaling (Supplemental Figure S5) and eight genes linked to nongenomic signaling (Supplemental Figure S6) were investigated. All 15 genes were expressed across all cohorts. There was evidence of down-regulation for four of the seven genomic signaling genes with anti-hormonal exposure in whole mammary tissue in the mouse models as these networks are consistent with activity at the transcriptional level. Nongenomic mechanisms are largely mediated by protein-protein interactions and were not expected to show major transcriptional changes.

Figure 5.

RNA-sequencing analyses indicating mouse estrogen receptor 1 (Esr1) and human cytochrome P450 family 19 subfamily A member 1 (aromatase; CYP19A1) mice with the highest numbers of differentially expressed genes (DEGs) at age 20 months (m), and Esr1 mice with significant enrichment in HALLMARK gene sets related to cell cycle progression. A: Bar graphs presenting numbers of statistically significantly differentially expressed protein-coding genes between Esr1 and CYP19A1 mice at age 18m. P-value adjusted (Padj) < 0.05 (DESeq2). B: Bar graphs showing three HALLMARK gene sets with smallest –LOG10 false discovery rate (FDR) q-values following Gene Set Enrichment Analysis (GSEA) of protein-coding genes significantly up-regulated in Esr1 mice at age 18m. C: Bar graphs showing three HALLMARK gene sets with smallest –LOG10 FDR q-values following GSEA of protein-coding genes significantly up-regulated in CYP19A1 mice at age 18m. D: Bar graphs presenting numbers of statistically significantly differentially expressed protein-coding genes between Esr1 and CYP19A1 mice at age 20m. Padj < 0.05 (DESeq2). E: Bar graphs showing three HALLMARK gene sets with smallest –LOG10 FDR q-values following GSEA of protein-coding genes significantly up-regulated in Esr1 mice at age 20m. F: Bar graphs showing three HALLMARK gene sets with smallest –LOG10 FDR q-values following GSEA of protein-coding genes significantly up-regulated in CYP19A1 mice at age 20m. G: Bar graphs presenting numbers of statistically significantly differentially expressed protein-coding genes between Esr1 and CYP19A1 mice at age 20m following 2 months tamoxifen exposure. Padj < 0.05 (DESeq2). H: Bar graphs showing three HALLMARK gene sets with smallest –LOG10 FDR q-values following GSEA of protein-coding genes significantly up-regulated in Esr1 mice at age 20m following 2 months tamoxifen exposure. I: Bar graphs showing three HALLMARK gene sets with smallest –LOG10 FDR q-values following GSEA of protein-coding genes significantly up-regulated in CYP19A1 mice at age 20m following 2 months tamoxifen exposure. J: Bar graphs presenting numbers of statistically significantly differentially expressed protein-coding genes between Esr1 and CYP19A1 mice at age 20m following 2 months letrozole exposure. Padj < 0.05 (DESeq2). K: Bar graphs showing three HALLMARK gene sets with smallest –LOG10 FDR q-values following GSEA of protein-coding genes significantly up-regulated in Esr1 mice at age 20m following 2 months tamoxifen exposure. L: Bar graphs showing three HALLMARK gene sets with smallest –LOG10 FDR q-values following GSEA of protein-coding genes significantly up-regulated in CYP19A1 mice at age 20m following 2 months tamoxifen exposure. n = 3 mice per cohort (A–L). E2, estrogen; E2F, E2 transcription factor; EMT, epithelial-mesenchymal transition; IFN, interferon; KRAS, Kirsten rat sarcoma viral oncogene homolog.

Figure 6.

Significant differences in expression levels of cell proliferation and estrogen response genes between mouse estrogen receptor 1 (Esr1) and human cytochrome P450 family 19 subfamily A member 1 (aromatase; CYP19A1) mice at age 20 months (m) were generally resolved by anti-hormonal exposure. Bar graphs showing comparative expression levels of cell proliferation and estrogen signaling genes from the human Prediction Analysis of Microarray 50 (PAM50) prognostic gene panel for estrogen receptor α–positive (ER⁺) breast cancer in Esr1 and CYP19A1 mice at age 18m (A), age 20m (B), age 20m following 2 months of tamoxifen exposure (C), and age 20m following 2 months of letrozole exposure (D). Data are given as means ± SEM (A–D). n = 3 mice per cohort (A–D). ∗P-value adjusted < 0.05 (DESeq2). Birc5, baculoviral IAP repeat containing 5; Ccnb1, cyclin B1; Cdc20, cell division cycle 20; Cdc6, cell division cycle 6; Cenpf, centromere protein F; Cep55, centrosomal protein 55; Exo1, exonuclease 1; Foxa1, forkhead box A1; Foxc1, forkhead box C1; Kif2c, kinesin family member 2; Krt14, keratin 14; Krt5, keratin 5; Mapt, microtubule-associated protein tau; Melk, maternal embryonic leucine zipper kinase; Mki67, marker of proliferation Ki-67; Mmp11, matrix metallopeptidase 11; Mybl2, MYB proto-oncogene like 2; Myc, MYC proto-oncogene, BHLH transcription factor; Pgr, progesterone receptor; Pttg1, PTTG1 regulator of sister chromatid separation, securin; Rrm2, ribonucleotide reductase regulatory subunit M2; Sfrp1, secreted frizzled related protein 1; TPM, transcripts per million; Tyms, thymidylate synthetase.

Aberrant Pregnancy Development-Related Gene Signature Identified in Esr1 Mice

Inspection of the cell proliferation– and estrogen response–related genes expressed at significantly higher levels in Esr1 mice that were coincidently members of the PAM50 gene panel revealed nine genes that are known to be developmentally regulated during pregnancy in the mammary gland (Figure 7A). The six cell proliferation–related genes were all up-regulated mid pregnancy (day 13, mouse pregnancy, GSE70440, https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE70440, last accessed September 17, 2022) with persistent higher expression levels through late pregnancy (day 18, mouse pregnancy, GSE70440, https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE70440, last accessed September 17, 2022). The three estrogen response–related genes were all down-regulated with late pregnancy. A heat map of the Esr1 and CYP19A1 mice focusing only on these nine genes demonstrated aberrant day 13 to 18 pregnancy-like patterns of expression in 18- and 20-month–old Esr1 mice and 18-month–old CYP19A1 mice with anti-hormonal exposure generally associated with the lowest expression levels (Figure 7B). To test whether other pregnancy development-related genes would show similar patterns, 34 additional genes known to be regulated during pregnancy were arrayed in heat maps for both the normal pregnancy reference data (Figure 7C) and the experimental cohorts (Figure 7D). The pattern persisted with genes expressed at higher levels mid and late pregnancy being higher expressed in CYP19A1 mice at age 18 months and Esr1 mice at ages 18 and 20 months. Of the 34 known pregnancy-related genes, 25 of them were identified as significant DEGs between the two models at 18 and/or 20 months of age. Taken together, the study of pregnancy-related gene expression in the two models revealed that Esr1 mice exhibited an abnormal mid-late pregnancy-like pattern of gene expression during reproductive senescence at age 20 months compared with CYP19A1 mice that more closely resemble normal non-pregnant patterns of gene expression at that end point.

Figure 7.

Mouse estrogen receptor 1 (Esr1) overexpression associated with abnormal expression patterns of pregnancy-related genes. A: Heat map showing relative expression levels of nine pregnancy-regulated genes identified as differentially expressed between Esr1 and human cytochrome P450 family 19 subfamily A member 1 (aromatase; CYP19A1) mice age 20 months (m) that are also members of the human Prediction Analysis of Microarray 50 (PAM50) prognostic gene panel for estrogen receptor α–positive (ER⁺) breast cancer at late pregnancy (day 18), mid-pregnancy (day 13), and in non-pregnant mice (downloaded data from 2-month–old mice: GSE70440, https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE70440, last accessed September 17, 2022). B: Heat map highlighting end points with aberrant pregnancy-like and non–pregnant-related expression patterns of the nine pregnancy-related genes across all four end points in the Esr1 and CYP19A1 mice. ¹Genes expressed at significantly higher levels in CYP19A1 mice age 20m: Esr1, matrix metallopeptidase 11 (Mmp11), and progesterone receptor (Pgr). ²Genes expressed at significantly higher levels in Esr1 mice age 20m: baculoviral IAP repeat containing 5 (Birc5), cyclin B1 (Ccnb1), keratin 14 (Krt14), keratin 5 (Krt5), marker of proliferation Ki-67 (Mki67), and MYB proto-oncogene like 2 (Mybl2). ³Krt14, Krt5, P-value adjusted (Padj) < 0.05, DESeq2, higher Esr1 age 20m with 2 months letrozole exposure. ⁴Krt5, Padj < 0.05, DESeq2, higher Esr1 age 20m with 2 months tamoxifen exposure. C: Heat map illustrating relative expression levels of an additional 34 pregnancy-regulated genes at late pregnancy (day 18), mid-pregnancy (day 13), and in non-pregnant mice. D: Heat map highlighting end points with aberrant pregnancy-like and non–pregnant-related expression patterns of the 34 additional pregnancy-related genes across all four end points in the Esr1 and CYP19A1 mice. ¹Genes expressed at significantly higher levels in CYP19A1 mice age 20m: Pleckstrin homology-like domain family A member 3 (Phlda3), AKT serine/threonine kinase 2 (Akt2), mitogen-activated protein kinase activated protein kinase 2 (Mapkapk2), transforming growth factor-β receptor 3 (Tgfbr3), cadherin 2 (Cdh2), transforming growth factor-β receptor 2 (Tgfbr2), bone morphogenetic protein 1 (Bmp1), cyclin G1 (Ccng1), and Polo-like kinase 3 (Plk3), Padj < 0.05, DESeq2. ²Genes expressed at significantly higher levels in Esr1 mice age 20m: growth differentiation factor 11 (Gdf11), BRCA1 DNA repair associated (Brca1), BRCA2 DNA repair associated (Brca2), proliferating cell nuclear antigen (Pcna), Aurora kinase A (Aurka), enhancer of zeste 2 polycomb repressive complex 2 subunit (Ezh2), cytokine-inducible SH2 containing protein (Cish), Tgfb3, BCL2-associated X, apoptosis regulator (Bax), SRY-box transcription factor 10 (Sox10), tumor necrosis factor receptor-associated factor 4 (Traf4), AKT serine/threonine kinase 1 (Akt1), Stat5a, transferrin receptor (Tfrc), E74-like ETS transcription factor 5 (Elf5), and casein α S1 (Csn1s1), Padj < 0.05, DESeq2. ³Phlda3, Tgfbr3 Padj < 0.05, DESeq2, higher CYP19A1 age 18m. ⁴Tgfb3, Sox10, Ccng1 Padj < 0.05, DESeq2, higher Esr1 age 18m. ⁵Gdf11 Padj < 0.05, DESeq2, higher Esr1 age 20m with both 2 months exposure to tamoxifen and letrozole. Yellow indicates highest expression level, and dark blue indicates lowest expression level, for each gene. Relative expression levels shown for unique individual samples. n = 3 for each cohort (B and D). Bcl6, BCL6 transcription repressor; Ccnd1, cyclin D1; Csn1s2a, casein α s2-like A; Csn1s2b, casein α s2-like B; Csn2, casein β; Csn3, casein κ; Mapk8, mitogen-activated protein kinase 8; Socs2, suppressor of cytokine signaling 2; Wap, whey acidic protein.

Pregnancy-Like Morphology Accompanies Pregnancy-Like Gene Expression with Resolution by Tamoxifen or Letrozole Exposure

Given the significant differences in mid-late pregnancy-related gene expression between the two models, the next question was whether mammary gland whole mount findings reminiscent of pregnancy would be found at higher prevalence in Esr1 compared with CYP19A1 mice. Mammary gland whole mounts were assessed for the appearance of dense uniform alveolar-like growth normally found during late pregnancy in mice as well as overall mammary gland density. Significantly higher prevalence of dense uniform alveolar growth and increased mammary density were identified in Esr1 compared with CYP19A1 mice at age 18 months (Figure 8, A and B). Tamoxifen exposure was associated with significant reductions in both alveolar-like growth and density, consistent with the significant reductions in tertiary branching and lobular growth, cellularity, HANs, preneoplasia, and cancer documented (Figures 2, A and B, and 3). This was a provocative finding because 12-month–old Esr1 mice experiencing transgene overexpression since birth demonstrate tamoxifen resistance linked to increased gene expression of interferon-γ–related genes.¹⁹ To test whether sensitivity to tamoxifen in the Esr1 mice undergoing reproductive senescence was associated with decreased expression levels of interferon-γ genes, relative gene expression levels were compared with previously published reference data (GSE63857, https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE63857, last accessed September 17, 2022) after renormalization with the current RNA-sequencing data set. For most genes identified as DEGs between Esr1 and CYP19A1 mice at age 12 months, expression was decreased in the Esr1 mice at ages 18 and 20 months (Figure 8C), consistent with a role for these genes in mediating tamoxifen resistance.

Figure 8.

Higher prevalence of dense uniform pregnancy-like alveolar growth and increased overall mammary gland density in mouse estrogen receptor 1 (Esr1) mice is resolved by tamoxifen exposure. A: Bar graphs illustrate percentage of mice with dense uniform pregnancy-like alveolar development in Esr1 and human cytochrome P450 family 19 subfamily A member 1 (aromatase; CYP19A1) cohorts. Esr1: age 18 months (m) n = 20; age 20m n = 19; age 20m tamoxifen n = 24; age 20m letrozole n = 20. CYP19A1: age 18m n = 21; age 20m n = 18; age 20m tamoxifen n = 23; age 20m letrozole n = 24. P values determined by Fisher exact test, two sided, GraphPad Prism version 9.3.1. B: Scatterplots illustrate distribution of relative mean mammary gland density scores in each cohort of Esr1 and CYP19A1 mice. Esr1: age 18m n = 18; age 20m n = 13; age 20m tamoxifen n = 15; age 20m letrozole n = 9. CYP19A1: age 18m n = 17; age 20m n = 14; age 20m tamoxifen n = 16; age 20m letrozole n = 16. Median indicated. P values determined by U-test, two tailed, GraphPad Prism version 9.3.1. Black arrow indicates that lower mammary gland density scores, which are based on pixel intensity readings of mammary gland whole mount images, correlate with higher mammary gland density. C: Heat map illustrating expression patterns of immune-related genes significantly differentially expressed between Esr1 and CYP19A1 mice at 12 months with transgene expression from birth [P-value adjusted (Padj) < 0.05, DESeq2] versus 18- and 20-month–old cohorts with transgene expression initiated at age 12 months. Differentially expressed genes (DEGs; Padj < 0.05, DESeq2) higher in Esr1 compared with CYP19A1 mice at age 18 m¹, age 20 m², age 20m following 2 months tamoxifen exposure³, and age 20m following 2 months of letrozole exposure⁴. DEGs (Padj < 0.05, DESeq2) higher in CYP19A1 compared with Esr1 mice at age 20 m⁵ and age 20m following 2 months of letrozole exposure⁶. Yellow indicates highest expression level, and dark blue indicates lowest expression level, for each gene. Relative expression levels shown for unique individual samples. 12m cohorts: n = 2. 18m and 20m cohorts: n = 3. ∗P < 0.05, ∗∗P < 0.01, and ∗∗∗P < 0.001. Dio2, iodothyronine deiodinase; Ear2, eosinophil-associated, ribonuclease A family, member 2; Fosb, FosB proto-oncogene, AP-1 transcription factor subunit; Gbp8, guanylate-binding protein 8; Gbp9, guanylate-binding protein 9; Gzmb, granzyme B; H2-Q7, histocompatibility 2, Q region locus 7; IFNG, interferon-γ; Il2rb, IL-2 receptor subunit β; Irf7, interferon regulatory factor 7; Irs2, insulin receptor substrate 2; Kbtdb12, Kelch repeat and BTB domain containing 12; NFKB, NF-κB; Nlrc5, NLR family CARD domain-containing 5; Nr4a1, nuclear receptor subfamily 4 group A member 1; Per1, period circadian regulator 1; Sectm1b, secreted and transmembrane protein 1b; Tap1, transporter 1, ATP binding cassette subfamily B member; Tfcp2l1, transcription factor CP2-like 1; Tgtp2, T-cell–specific GTPase 2; TNFA, tumor necrosis factor-α; Wnk4, WNK lysine-deficient protein kinase 4; Zbp1, Z-DNA binding protein 1.

Discussion

A fundamental question in biology and medicine and especially in age-associated diseases, such as breast cancer, is how aging influences disease presentation.⁷⁴ In women, age-specific rates of ER⁺ breast cancers increase with age following menopause.⁷⁵ Overexpressions of Esr1 and CYP19A1 in breast tissue are two distinct perturbations of estrogen pathway signaling found in women with breast cancer,^5,6,76, 77, 78, 79 that are modeled in the mice presented herein.^19,34,80 This study demonstrated that disease presentation in the Esr1 and CYP19A1 mice is impacted by age with accompanying reproductive senescence. Although pre–reproductive senescent 12-month–old mice show slightly higher disease rates with CYP19A1 overexpression compared with Esr1 overexpression, this was reversed in 20-month–old mice in reproductive senescence, where Esr1 overexpression was associated with significantly higher disease rates. This higher disease rate was linked to a pattern of higher expression of proliferation-related genes that was maintained through reproductive senescence, even as estrogen-linked ovarian function was lost and there was evidence of mammary gland involution noted by loss of tertiary branching, lobular growth, and overall mammary gland density. This pattern was not seen in the CYP19A1 mice, showing that ovarian reproductive senescence impacted these two components of the estrogen signaling pathway differently. A possible explanation for the lower rates of cell proliferation and disease in the CYP19A1 mice could be decreased androgen production, required by aromatase for estrogen production, from the senescent ovaries.⁸¹

Persistent activity of ER α–driven growth pathways through reproductive senescence maintained an aberrant pregnancy-like proliferative gene expression profile in the Esr1 mice that correlated with a higher prevalence of preneoplasia and cancer in this model. High-risk breast cells from women with breast cancer can exhibit a similar profile.²⁸ During normal pregnancy, estrogen signaling pathways initiate a stage of proliferative alveolar morphogenesis that is followed by differentiation into milk-producing cells.⁸² Failure of proliferative cells to differentiate is a cancer risk factor, well studied in human papillomavirus–induced cervical carcinogenesis.⁸³ Pregnancy at younger ages exerts a protective effect on breast cancer generation, but initial pregnancy at older ages is associated with increased breast cancer risk.84, 85, 86 In this study, induction of abnormal levels of Esr1 was initiated in 12-month–old mice, equivalent to human middle age. The significant amount of preneoplasia and cancer found in both 18- and 20-month–old Esr1 mice only 6 and 8 months following transgene expression may be attributed to the aging mammary epithelium that the transgene induction was initiated in.

Both tamoxifen and letrozole exposure were effective in reducing prevalence of preneoplasia and cancer. This was an unexpected finding in the Esr1 mice as they have reproducibly demonstrated tamoxifen resistance linked to the presence of an activated Irf7-Stat1 network.^19,35,36 The presence of tamoxifen responsiveness demonstrated herein was associated with down-regulation of this network. Similar to this observation, a second immunologic pathway, NF-κB subunit 1 signaling, has been linked to tamoxifen resistance in women⁸⁷ and was activated in the younger tamoxifen-resistant Esr1 mice but down-regulated in the tamoxifen-responsive older mice. Most preneoplastic and cancer disease found in the mice was ER α⁺, consistent with its resolution with tamoxifen and letrozole exposure.

In summary, investigation of mammary disease initiated by overexpression of Esr1 and CYP19A1 in mid-age mice yielded a series of new observations relevant to human breast cancer. First, Esr1 overexpression but not CYP19A1 overexpression was associated with a persistent proliferative response through reproductive senescence. Second, this proliferative response was embedded within an aberrant pregnancy-like gene expression profile and included known prognostic signature genes for human ER⁺ breast cancer. Third, aging and entry into reproductive senescence altered tamoxifen responsiveness of the Esr1 mice. Finally, expansion of mouse model studies of breast cancer generation through natural reproductive senescence yielded new insights into mammary disease induced by Esr1 and CYP19A1 overexpression that more closely resembles the natural history of breast carcinogenesis in women, where most of the disease develops with and after menopause.

Author Contributions

P.A.F. conceived the study; P.A.F., W.W., and J.A.F. designed the study; P.A.F., W.W., K.K., B.L.R., V.M., X.Z., and J.A.F. acquired data; P.A.F., K.K., B.L.R., G.K., V.M., and J.A.F. interpreted and analyzed data; P.A.F. and B.L.R. prepared the initial manuscript; P.A.F., W.W., K.K., B.L.R., G.K., V.M., X.Z., and J.A.F. reviewed the manuscript before submission and acceptance of authorship; and P.A.F. secured funding. P.A.F. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Supplemental Data

Supplemental Figure S1.

Transgene expression validation and representative images of estrogen receptor α (ER) immunohistochemistry of mammary gland ducts from cohorts not shown in Figure 3, E–G. A: Bar graphs illustrate relative percentage transgene expression levels in mammary tissue from each cohort compared with age 18 months (m; set at 100%). Two unique primer pairs (Sp3 and Sp4) are presented for mouse estrogen receptor 1 (Esr1) mice, and one primer pair is presented for human cytochrome P450 family 19 subfamily A member 1 (aromatase; CYP19A1) mice. Esr1: age 18m n = 2; age 20m n = 2; age 20m tamoxifen n = 3; age 20m letrozole n = 1. CYP19A1: age 18m n = 3; age 20m n = 1; age 20m tamoxifen n = 1; age 20m letrozole n = 1. B–F: ER immunohistochemistry of ducts from 18-month–old Esr1 (B), 18-month–old CYP19A1 (C), 20-month–old CYP19A1 (D), 20-month–old CYP19A1 with 2 months tamoxifen exposure (E), and 20-month–old Esr1 with 2 months letrozole exposure mice (F). Black arrows indicate representative cells with nuclear-localized ER staining. All images scaled identically. Scale bar = 10 μm (B–F). Original magnification, ×40 (B–F).

Supplemental Figure S2.

Most cell proliferation genes from the Prediction Analysis of Microarray 50 (PAM50) panel that were significantly differentially expressed between mouse estrogen receptor 1 (Esr1) and human cytochrome P450 family 19 subfamily A member 1 (aromatase; CYP19A1) mice at age 20 months (m) were also significantly down-regulated with tamoxifen and letrozole in Esr1 mice. Bar graphs illustrating relative transcripts per million (TPM) in each cohort for each gene expressed at significantly different levels between Esr1 and CYP19A1 mice at age 20 months (Figure 6B). Significant differences between cohorts within each genotype are indicated. Relative TPM values shown for baculoviral IAP repeat containing 5 (Birc5; A), cyclin B1 (Ccnb1; B), cell division cycle 20 (Cdc20; C), cell division cycle 6 (Cdc6; D), centromere protein F (Cenpf; E), centrosomal protein 55 (Cep55; F), exonuclease 1 (Exo1; G), forkhead box C1 (Foxc1; H), kinesin family member 2 (Kif2ca; I), maternal embryonic leucine zipper kinase (Melk; J), marker of proliferation Ki-67 (Mki67; K), MYB proto-oncogene like 2 (Mybl2; L), MYC proto-oncogene, BHLH transcription factor (Myc; M), PTTG1 regulator of sister chromatid separation, securin (Pttg1; N), ribonucleotide reductase regulatory subunit M2 (Rrm2; O), and thymidylate synthetase (Tyms; P). Relative expression levels shown for unique individual samples. Data are given as means ± SEM (A–P). n = 3 for each cohort (A–P). ∗P-value adjusted (Padj) < 0.05, ∗∗Padj < 0.01, ∗∗∗Padj < 0.001, and ∗∗∗∗Padj < 0.0001 (DESeq2).

Supplemental Figure S3.

Estrogen response genes from Prediction Analysis of Microarray 50 (PAM50) panel showed variable behavior with exposure to tamoxifen and letrozole in both mouse estrogen receptor 1 (Esr1) and human cytochrome P450 family 19 subfamily A member 1 (aromatase; CYP19A1) mice. Bar graphs illustrating relative transcripts per million (TPM) in each cohort for each gene expressed at significantly different levels between Esr1 and CYP19A1 mice at age 20 months (Figure 6B). Significant differences between cohorts within each genotype indicated. Relative TPM values shown for Esr1 (A), forkhead box A1 (Foxa1; B), microtubule-associated protein tau (Mapt; C), matrix metallopeptidase 11 (Mmp11; D), progesterone receptor (Pgr; E), and secreted frizzled-related protein 1 (Sfrp1; F). Relative expression levels shown for unique individual samples. Data are given as means ± SEM (A–F). n = 3 for each cohort (A–F). ∗P-value adjusted (Padj) < 0.05, ∗∗Padj < 0.01, and ∗∗∗∗Padj < 0.0001 (DESeq2).

Supplemental Figure S4.

Relative expression levels of luminal and basal cytokeratins across cohorts. Bar graphs illustrating relative transcripts per million (TPM) in each cohort for keratin (Krt) 7 (A), Krt8 (B), Krt18 (C), Krt19 (D), Krt5 (E), and Krt14 (F). Relative expression levels shown for unique individual samples. Data are given as means ± SEM (A–F). n = 3 for each cohort (A–F). ∗P-value adjusted (Padj) < 0.05, ∗∗Padj < 0.01, ∗∗∗Padj < 0.001, and ∗∗∗∗Padj < 0.0001 (DESeq2). CYP19A1, human cytochrome P450 family 19 subfamily A member 1 (aromatase); Esr1, mouse estrogen receptor 1; m, months.

Supplemental Figure S5.

Selected genes previously linked to estrogen receptor α (ER) genomic signaling showed significant down-regulation following exposure to an anti-hormonal. Bar graphs illustrating relative transcripts per million (TPM) in each cohort for genes associated with ER genomic signaling. Significant differences between cohorts within each genotype indicated. Relative TPM values shown for MYC proto-oncogene, BHLH transcription factor (Myc; A), cyclin D1 (Ccnd1; B), cyclin-dependent kinase 2 (Cdk2; C), bcl2-like 1 (Bcl2l1; D), cyclin E1 (Ccne1; E), cyclin-dependent kinase inhibitor 1a (Cdkn1a; F), and Bcl2 apoptosis regulator (Bcl2; G). Relative expression levels shown for unique individual samples. Data are given as means ± SEM (A–G). n = 3 for each cohort (A–G). ∗P-value adjusted (Padj) < 0.05, ∗∗Padj < 0.01, ∗∗∗Padj <0.001, and ∗∗∗∗Padj < 0.0001 (DESeq2). CYP19A1, human cytochrome P450 family 19 subfamily A member 1 (aromatase); Esr1, mouse estrogen receptor 1; m, months.

Supplemental Figure S6.

Selected genes previously linked to estrogen receptor α (ER) nongenomic signaling were expressed in mammary tissue but not generally down-regulated following exposure to an anti-hormonal. Bar graphs illustrating relative transcripts per million (TPM) in each cohort for genes associated with ER genomic signaling. Significant differences between cohorts within each genotype indicated. Relative TPM values shown for insulin growth factor receptor 1 (Igf1r; A), matrix metallopeptidase 2 (Mmp2; B), matrix metallopeptidase 9 (Mmp9; C), epidermal growth factor receptor (Egfr; D), mitogen-activated protein kinase 1 (Mapk1; E), SHC adaptor protein 1 (Shc1; F), growth factor receptor-bound protein 2 (Grb2; G), and SOS Ras/Rac guanine nucleotide exchange factor 1 (Sos1; H). Relative expression levels shown for unique individual samples. Data are given as means ± SEM (A–H). n = 3 for each cohort (A–H). ∗P-value adjusted (Padj) < 0.05, ∗∗∗∗Padj < 0.0001 (DESeq2). CYP19A1, human cytochrome P450 family 19 subfamily A member 1 (aromatase); Esr1, mouse estrogen receptor 1; m, months.