Responsiveness of Brca1 and Trp53 Deficiency–Induced Mammary Preneoplasia to Selective Estrogen Modulators versus an Aromatase Inhibitor in Mus musculus

Research into breast cancer prevention in the setting of Breast cancer 1 (BRCA1) mutation continues due to considerations of efficacy and side-effects of current approaches (1–3), in conjunction with the possibility of developing more effective chemopreventive methods in the future (4–7). Tamoxifen (1, 2, 8, 9), raloxifene (8, 10), and letrozole (11) have all been proposed as risk-reducing antihormonals for women carrying germline BRCA1 mutations. Conditional genetically engineered mouse models (GEMM) where both Brca1 alleles are disrupted in mammary epithelial cells (MEC) coupled with Tumor protein P53 (Trp53) haploinsufficiency have been recurrently employed to study pathophysiology, prevention, and treatment of mammary cancer related to loss of Brca1 function (5–7, 12, 13). Mammary cancers in these models are predominantly triple negative and demonstrate a range of histologies (5, 12–14). To model the situation in women, where only one BRCA1 allele is mutated (1, 2), here we characterized and compared mammary cancer and preneoplasia pathophysiology and preventive approaches in GEMM with one (Brca1^{wildtype (WT)/floxed (fl11)/Cre/p53−/+}) versus two Brca1 (Brca1^{floxed (fl)11/fl11/Cre/p53−/+}) alleles disrupted in combination with Trp53 haploinsufficiency. Mammary cancer development in mice is typically preceded by the appearance of preneoplastic lesions including hyperplastic alveolar nodules (HAN), a measure that can be reliably and reproducibly quantified on mammary gland whole mounts (5, 13–16). In women, the presence of mammary epithelial cell cytologic atypia has been discussed as an indicator of breast cancer risk (17).

Loss or decreased TRP53 function has been proposed as an early event in Brca1 mutation-related breast cancer progression with BRCA1 loss of heterozygosity (LOH) occurring later and perhaps not required for carcinogenesis (18). Trp53 heterozygosity has been observed to be obligatory for mammary cancer generation in GEMM with both one and two Brca1 alleles disrupted (12, 19). Loss of PTEN (18), pathogenic mutations in Death-associated protein kinase 3 (DAPK3), Transmembrane protein 135 (TMEM135), and GATA binding protein 4 (GATA4; ref. 20), and kindreds showing mutation in both BRCA1 and Breast cancer 2 (BRCA2; ref. 21) have been reported to contribute to human breast cancer generation. Trp53 LOH has been reported in mammary cancer cell cultures derived from Brca1^{fl11/fl11/Cre/p53−/+} mice (22).

In the mammary gland, hormonal signaling stimulates tertiary branching and MEC proliferation (23, 24). Tamoxifen is reported to reduce levels of MEC proliferation (25). Loss of BRCA1 can enhance estrogen and progesterone signaling in human breast cells (4, 26, 27) and GEMM with disrupted Brca1 alleles also demonstrate abnormal growth responses to estrogen and/or progesterone pathway stimulation (14, 23, 28). Loss of BRCA1 in cancer cells can interrupt the inhibitory action of tamoxifen on estrogen signaling (13, 29). Tamoxifen exposure promoted mammary cancer development in Brca1^{fl11/fl11/Cre/p53−/+} mice in a previous study (13). Here we specifically examined whether or not loss of Brca1 or any of the other genetic components of the Brca1^{fl11/fl11/Cre/p53−/+} mice interrupted the inhibitory action of tamoxifen in mammary tissue and compared the impact of tamoxifen to the alternative selective estrogen modulator raloxifene and the nonsteroidal aromatase inhibitor letrozole.

Nulliparous C57Bl/6 female mice carrying different combinations of genetically engineered modifications in Brca1 (30), Trp53 (31), the MMTV-Cre D transgene, which targeted loss of full-length Brca1 to MEC (32), or no genetic modifications (wild-type: WT) were generated in breeding colonies at Georgetown University (Washington, DC). Control and treatment mice for each genotype were derived from the same genetic crosses. Brca1^{WT/fl11/Cre/p53−/+}, Brca1^{fl11/fl11/Cre}, Brca1^WT/fl11/Cre, Trp53 (p53)−/+, and MMTV-Cre (Cre) mice were derived from Brca1^{fl11/fl11/Cre/p53−/+} mice, maintained since 2002 on a C57Bl/6 background, and bred to C57Bl/6 WT mice from our breeding colony. Strain 129 embryonic stem cells were used for Brca1^fl11(NIH) and disrupted Trp53 (AB1) allele generation (NIH; refs. 30, 31). Mice were maintained in barrier zones, weaned (age 3 weeks), genotyped (Transnetyx Inc.), allocated in turn to different observation/treatment groups for inclusion in the study based on genotype, and maintained in single-sex sterilized ventilated cages with corncob bedding (1–5 mice per cage) with ad libitum access to water/chow under 12-hour dark/light cycles at Georgetown University (Washington, DC). Genotypes demonstrated equivalent weights, growth, activity, and fertility. Observation groups (Ob) were aged to 12 months for appearance of tumor. Treatment/control groups were aged to 4 months for tamoxifen (T; 25 mg/60 day/subcutaneous pellet_12.99 mg/kg/day, Innovative Research of America; refs. 13, 15), raloxifene (R; 22.5 mg/60 day/subcutaneous pellet_11.44 mg/kg/day, Innovative Research of America; refs. 33, 34), letrozole (L; 2.5 mg/60 day/subcutaneous pellet_1.29 mg/kg/day, Innovative Research of America; ref. 15), placebo pellet (P; Innovative Research of America), or sham surgery. Initial group size was determined from previous experiments (5, 13, 14) to ensure n ≥ 10 at observation/treatment endpoints. All mice were monitored bi-weekly for health and tumor development until euthanasia and necropsy at either age 6 or 12 months or when largest tumor reached diameter of approximately 2 cm (5, 13, 14). Five mice were excluded from the study for health reasons (open wound/pellet placement site n = 4, ear infection n = 1). There were no significant differences in mouse weights at study endpoints between treatment and control (one-way ANOVA, Sidak multiple comparisons test) or observation (unpaired t test) groups (P < 0.05, GraphPad Prism version 6.01). Genotype numbers (n)/observation/treatment/control group and [mean weights (grams) ± SEM/age in months (m)]: Brca1^{fl11/fl11/Cre/p53−/+} (n = 71: Ob: n = 10, T: n = 16, R: n = 11, L: n = 11, P: n = 12, S: n = 11; 39.2 ± 2.0/12m, 33.2 ± 0.8/6m), Brca1^{WT/fl11/Cre/p53−/+}(n = 70: Ob: n = 12, T: n = 12, R: n = 11, L: n = 12, P: n = 12, S: n = 11; 34.4 ± 5.0/12m, 33.5 ± 1.3/6m), Brca1^{fl11/fl11/Cre} (n = 25: T: n = 12, S: n = 13; 33.7 ± 1.7/6m), Brca1^WT/fl11/Cre (n = 20: T: n = 10, S: n = 10; 30.9 ± 0.8/6m), p53−/+ (n = 20: T: n = 10, S: n = 10; 28.0 ± 0.6/6m), Cre (n = 23: T: n = 11, S: n = 12; 28.0 ± 1.5/6m), WT (n = 22: T: n = 10; S: n = 12; 32 ± 1.0/6m). Georgetown University Animal Care and Use Committee (Protocols 15-020-100220/12-020-100034) and Institutional Biosafety Committee approved all procedures.

HAN numbers per inguinal mammary gland (5, 13–16) and presence of tertiary branching (24) were determined blindly by three independent observers (S.J. Alothman, W. Wang, and P.A. Furth) on mammary gland whole mounts fixed in Carnoy solution and stained in carmine alum. The mean of the three readings was calculated for numerical data and the two concordant readings used for discrepant categorical data. Cancer histology and presence or absence of cytologic atypia and nuclear localized Estrogen Receptor (ER) alpha in MEC were read blindly on hematoxylin and eosin (H&E)-stained sections of formalin-fixed, paraffin-embedded (FFPE) tissue by a board-certified pathologist (B.V. Kallakury). Cytologic atypia was defined as the presence of mild, moderate, or severe abnormal nuclear architecture. IHC was performed on FFPE tissue following antigen retrieval using antibodies directed against proliferating cell nuclear antigen (PCNA; ref.13; 1:300 1-hour room temperature, sc- 7907, Santa Cruz Biotechnology) and ER alpha (refs. 5, 15; 1:225, 1-hour room temperature, sc-542, Santa Cruz Biotechnology). To calculate proliferative indices (PI), MEC percentages demonstrating PCNA nuclear localization within a total population of 500–1,000 MECs were determined blindly by two independent observers (S.J. Alothman and P. A. Furth) and means presented. All available samples were analyzed. N for each treatment/control genotype group included in figure legends. H&E/IHC sections were not available from all mice in some groups due to inadequate fixation technique or expended sample material. Digital photographs were taken using a Nikon Eclipse E800M microscope and DMX1200 camera with the NIS Elements imaging software version 4.30 program (Nikon Corporation).

DNA samples were prepared for custom DNA capture and sequencing using the SureSelect Target Capture Enrichment System (Agilent Technologies). DNA was isolated from cell culture and tissue sample lysates from Brca1^{fl11/fl11/Cre/p53−/+} and Brca1^{WT/fl11/Cre/p53−/+} mice using the MasterPureComplete DNA Purification Kit (Epicentre, Illumina Inc.). Custom cRNA baits for the genes of interest (Brca1, Trp53, Brca2, Dapk3, Pten, and Tmem135) were designed using SureDesign software (Agilent Technologies). The genomic coordinates of each gene were derived from the Mus musculus NCBI build 37/UCSC mm9 assembly (MGSCv37, July 2007), and covered the entire transcribed region, including coding exons, introns, 5′ and 3′ untranslated regions, and an additional 10 kb total of contiguous sequence from the 5′ and 3′ ends for each gene. The genomic coordinates for each targeted gene were as follows: Brca1 (chr11:101345078101418269), Trp53 (chr11:69388861-69410375), Brca2 (chr5:151320198-151377324), Dapk3 (chr10:80641008-80660942), Pten (chr19:32827067-32905650), and Tmem135 (chr7:96283231-96492297). The region size of the target sequences totaled 4,59,420 kb over 6,341 cRNA probes (sequences available upon request) resulting in >90% coverage. Libraries were prepared for sequencing using the SureSelectXT Custom 1Kb-499kb Library Kit and SureSelectXT Rreagent Kit (Agilent Technologies). The custom capture DNA libraries were then sequenced on the Illumina MiSeq platform (paired-end 2 × 150 bp) using MiSeq Reagent Kit V2 (Illumina Inc.) with sequences were uploaded to Sequence Read Archive (SRA) BioProject ID PRJNA378147. Sequence reads were processed with a pipeline consisting of the following elements: (i) base calls generated in real-time on the Agilent SureSelect system; (ii) Perl scripts developed in-house to produce demultiplexed FASTQ files by lane and index sequence; (iii) demultiplexed BAM files aligned to the Mus musculus 9 reference genome using BurrowsWheeler Aligner (35). Read-pairs not mapping within +2 SDs of the average library size (*125 + 15 bp for exomes) were removed. All aligned read data were subjected to the following steps: (i) duplicate removal was performed, (i.e., the removal of reads with duplicate start positions; Picard MarkDuplicates, https://broadinstitute.github.io/picard/); (ii) indel realignment was performed (GATK IndelRealigner, https://software.broadinstitute.org/gatk/gatkdocs/org_broadinstitute_gatk_tools_walkers_indels_IndelRealigner.php) resulting in improved base placement and lower false variant calls; (iii) base qualities were recalibrated (GATK Table Recalibration, https://software.broadinstitute.org/gatk/gatkdocs/org_broadinstitute_gatk_tools_walkers_bqsr_BaseRecalibrator.php). Variant detection and genotyping were performed using the UnifiedGenotyper tool (GATK (v3.5) https://software.broadinstitute.org/gatk/gatkdocs/org_broadinstitute_gatk_tools_walkers_genotyper_UnifiedGenotyper.php). Variant data for each sample were formatted (variant call format) as “raw” calls that contain individual genotype data for one or multiple samples, and flagged using the filtration walker (GATK) to mark sites that are of lower quality/false positives, for example, low quality scores (≤ 50), allelic imbalance (≥0.75), long homopolymer runs (> 3), and/or low quality by depth (QD< 20). SNP were visualized using the Integrative Genomics Viewer 2.3.80 (36, 37) to examine for LOH and mutation. Regions demonstrating deviation from the reference genome were identified and compared with regions that were modified in the Brca1 and Trp53 loci in the GEMM (30, 31). For these analyses, DNA was extracted directly from noncancerous mammary gland (MG) or cancer (CA) tissue or primary cultures of noncancerous or cancer MEC under conditional reprogramming culture without feeder cells (CRC) or EpiCult B Mouse Medium (EpiC; Stemcell Technologies; ref. 38). Primary MEC cultures were obtained from mammary glands of Brca1^{fl11/fl11/Cre/p53−/+} and Brca1^{WT/fl11/Cre/p53−/+} mice aged 8–11months (2011–2012), stored viably at −80°C when not passaged, authenticated (2015) for the presence of Brca1 exon 11 deletion and Cre transgene by RT-PCR and PCR (38), and verified as having predicted genetically engineered deviations from the reference genome in Brca1 and Trp53 genes in this study (2016). Mycoplasma testing has not been performed. Passage numbers (p) reported are from initial isolation (p0). In some cases, samples from a single mouse from noncancerous and cancer tissue and/or under both culture conditions included, indicated by the same superscript number [T, tamoxifen exposure; samples tested: Brca1^{fl11/fl11/Cre/p53−/+}; n = 7 (noncancerous: MG/6m, n = 1; CRCp5/11m, n = 1; cancer: CA/6m, n = 1; CA/6m/T, n = 1; CRCp12/11m, n = 1; CRCp8/8m, n = 1¹; EpiCp9/8m, n = 1¹); Brca1^{WT/fl11/Cre/p53−/+}, n = 6 (noncancerous: CRCp4/10m, n = 1²; EpiCp3/10m, n = 1²; cancer: CRCp7/10m, n = 1²; EpiCp17/10m, n = 1²; CRC/p12/11m, n = 1³; EpiCp11/1m, n = 1³; WT, n = 1 (noncancerous: MG/6m)].

The primary endpoints were mammary cancer and presence or absence of HANs. Secondary endpoints included mean age when mammary tumor development reached 2-cm diameter, HAN numbers/inguinal mammary gland, tertiary branching and proliferative index, and cytologic atypia. Statistical analyses were performed with GraphPad Prism version 6.01 using Fisher exact test for prevalence of cancer, HANs, tertiary branching, and cytologic atypia (P < 0.05) and Mann–Whitney test for proliferative indices and HAN number (P < 0.05). Percent change was calculated using the formula (B − A/A) × 100, where A was the value for the control group and B the value for the treatment group.

As women develop breast cancer when only one BRCA1 allele is mutated, we sought to determine whether the same could be true in mice to more closely model the impact of preventive regimens in mice with loss of only one Brca1 allele. Published work documents that disruption of Trp53 frequently occurs in breast cancers that develop in women carrying BRCA1 mutations (18), contributes to cancer generation in mice with disrupted Brca1 alleles (12, 14, 19, 30), and by itself does not lead to mammary cancer by age 12 months (16). Like Brca1^{fl11/fl11/Cre/p53−/+} mice, Brca1^{WT/fl11/Cre/p53−/+} mice developed invasive mammary cancer by age 12 months without significant difference in latency (Fig. 1A and B) and demonstrated HAN development on mammary gland whole mounts (Fig. 1C and D). Both GEMM showed a range of invasive mammary cancer histologic subtypes from more differentiated adenocarcinomas to less differentiated sarcomatoid carcinomas (Fig. 2A–H). HANs appeared similarly (Fig. 2I–J) and mammary epithelial cytologic atypia was found in both (Fig. 2K). Some, but not all atypical MEC demonstrated nuclear-localized ERα in both models (Fig. 2L). There was no significant difference in the percentages of mice from the two different genotypes that demonstrated tertiary branching (Fig. 2M).

We then tested for Brca1 or Trp53 LOH and mutation of Brca2, Dapk3, Pten, and Tmem135 in noncancerous and cancer tissue from the two GEMM and WT mice. Both intronic and exonic sequences were included in the analyses because pathogenic intronic mutations are reported in humans (39). There were no SNPs in either Brca1 or Trp53 uniquely identified in cancers from Brca1^{WT/fl11/Cre/p53−/+}mice as compared with those from Brca1^{fl11/fl11/Cre/p53−/+} mice. SNPs were identified in the Brca1 gene in Brca1^{fl11/fl11/Cre/p53−/+} and Brca1^{WT/fl11/Cre/p53−/+} mice but only in regions that were genetically engineered to introduce lox sites in the Brca1 gene (ref. 30; Fig. 3A). All but two of the cancers from both genotypes showed SNPs in the region of the genetically engineered Trp53 gene (ref. 31; Fig. 3B). SNPs outside this region were detected in one of the two Brca1^{fl11/fl11/Cre/p53−/+} cancers directly analyzed and in the EpiC, but not CRC, culture of noncancerous MEC from a Brca1^{WT/fl11/Cre/p53−/+} mouse. The few SNPs identified in Brca2 and Dapk3 genes were also present in the WT tissue analyzed (Fig. 3C and D). Analysis of Pten and Tmem135 gene sequences revealed multiple SNPs present in both GEMM and WT samples (data not shown).

In Brca1^{fl11/fl11/Cre/p53−/+} mice, tamoxifen exposure significantly reduced prevalence of tertiary branching by 65% and proliferative index (PI) by 60% (Fig. 4). In WT mice, tamoxifen exposure significantly reduced tertiary branching prevalence by 88% and in Brca1^WT/fl11/Cre mice by 50%, and PI was significantly reduced in Brca1^WT/fl11/Cre mice by 42%. While the magnitude of reduction by tamoxifen in tertiary branching and PI varied across different genotypes, no evidence was found to indicate that Brca1^{fl11/fl11/Cre/p53−/+} mice were less sensitive to the antihormonal action of tamoxifen than mice carrying more limited genetic modifications. Tamoxifen had no impact on HAN or cancer prevalence or HAN number in Brca1^{fl11/fl11/Cre/p53−/+} mice (Fig. 5A–E). In contrast, tamoxifen exposure did statistically significantly reduce HAN number in Brca1^{WT/fl11/Cre/p53−/+} mice by 95% (Fig. 5B). Overall cytologic atypia was detected less frequently in tamoxifen-exposed mice (Fig. 5F). Atypical MECs in tamoxifen-exposed Brca1^{fl11/fl11/Cre/p53−/+} mice retained nuclear-localized ERα expression (Fig. 5G–J).

Raloxifene statistically significantly reduced tertiary branching in Brca1^{fl11/fl11/Cre/p53−/+} mice by 71%, but was without significant impact on HAN prevalence or number (Fig. 6A and B). In contrast, letrozole statistically significantly reduced tertiary branching by 78%, HAN prevalence by 52%, and HAN number by 75% (Fig. 6A and B). In Brca1^{WT/fl11/Cre/p53−/+} mice, the impact of raloxifene and letrozole were relatively similar (Fig. 6C). Cancer appeared in control (Fig. 6D) but not raloxifene- or letrozole-exposed mice. Letrozole was more effective than either tamoxifen or raloxifene in reducing preneoplasia in Brca1^{fl11/fl11/Cre/p53−/+} mice but HANs were not entirely eradicated and cytologic atypia persisted. Similar results were found in Brca1^{WT/fl11/Cre/p53−/+}mice (Fig. 6E–V). Atypical MECs in letrozole- and raloxifene-exposed Brca1^{fl11/fl11/Cre/p53−/+} and Brca1^{WT/fl11/Cre/p53−/+}mice retained nuclear-localized ERα expression (insets, Fig. 6G, J, M, P, S, V).

Pharmacologic approaches to breast cancer prevention can be understood in degrees of risk reduction for a specific individual and agent weighed against possible side effects and potential harms (40). Currently, tamoxifen and raloxifene are recommended by the United States Preventive Services Task Force (41). The American Society of Clinical Oncology Clinical Practice Guidelines adds the possibility of an aromatase inhibitor (exemestane) for postmenopausal women (42). Both groups agree that current evidence fails to show any impact of antihormonal agents on prevention of ER-negative breast cancer or all-cause mortality. Women who carry BRCA1 mutations are of particular concern due to their defined genetic risk factor and their higher proportion of ER-negative breast cancers. Risk-reducing mastectomy and salpingo-oophorectomy offer the highest likelihood of reducing both cancer diagnosis and all-cause mortality (43).

One of the criticisms of the Brca1^{fl11/fl11/Cre/p53−/+}mouse model has been that both Brca1 alleles are disrupted while in women only one Brca1 allele is mutated. It has been suggested that MECs with only one disrupted BRCA1 allele may be more sensitive to antihormonal therapies than cells with two disrupted BRCA1 alleles (26). Here we tested the impact of antihormonals on preneoplasia development with both one and two Brca1 alleles and found evidence of increased responsiveness to tamoxifen in Brca1^{WT/fl11/Cre/p53−/+} as compared with Brca1^{fl11/fl11/Cre/p53−/+} mice. There has been significant discussion around whether or not Brca1 LOH is required for cancer development (18, 19). Significantly we determined that Brca1 LOH was not required for mammary cancer development and found no evidence that Trp53 LOH was required for cancer development even though this has been previously reported for this model (22).

Letrozole has been proposed as an alternative antihormonal in women carrying BRCA1 or BRCA2 mutations (11). We found letrozole to be more effective than either tamoxifen or raloxifene in reducing preneoplasia in Brca1^{fl11/fl11/Cre/p53−/+}mice. Because we did not follow cohorts of mice after antihormonal therapy was completed as was done previously (13), we cannot say whether or not the interim antihormonal therapy would have had a significant impact on later cancer generation. However, the persistence of HANs after antihormone exposure suggests that these mice remained at risk for cancer development even when statistically significant reductions in prevalence of preneoplasia was found. Retention of nuclear-localized ERα in cytologically atypical cells following antihormone exposure indicated that the therapy did not interrupt presence of this cell population. Whether or not these cells contribute to cancer generation in either this model or in women who develop triple-negative cancer remains an open question.

Women interested in pharmacologic approaches for breast cancer prevention look for approaches that are highly effective without significant side effects (40). Even if antihormone sensitivity is greater with loss of a single as compared with both Brca1 alleles, BRCA1 LOH is well documented in breast cancers developing in women who carry one mutated BRCA1 allele (18). It is prudent for investigators to seek chemopreventive approaches that would be effective under both genetic settings, as the precise genetic makeup of each breast epithelial cell in a population is not easily defined. Use of both Brca1^{WT/fl11/Cre/p53−/+} and Brca1^{fl11/fl11/Cre/p53−/+} mouse models in chemoprevention studies has the advantage of discerning how proposed agents may act when Brca1 is deficient as compared with only haploinsufficient.

In summary, these studies report that antihormonal agents can significantly impact mammary disease prevalence in mouse models of Brca1 mutation but fail to eradicate preneoplasia. While loss of one as compared with two Brca1 alleles may increase responsiveness to antihormonal agents, the call for more effective approaches persists.

No potential conflicts of interest were disclosed.

Conception and design: S.J. Alothman, W. Wang, P.A. Furth

Development of methodology: S.J. Alothman, W. Wang, D.S. Goerlitz, Md. Islam, A. Kishore, B. Kallakury, P.A. Furth

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.J. Alothman, W. Wang, D.S. Goerlitz, Md. Islam, A. Kishore, R.I. Azhar, B. Kallakury, P.A. Furth

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.J. Alothman, W. Wang, X. Zhong, A. Kishore, R.I. Azhar, B. Kallakury, P.A. Furth

Writing, review, and/or revision of the manuscript: S.J. Alothman, W. Wang, Md. Islam, X. Zhong, A. Kishore, B. Kallakury, P.A. Furth

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): W. Wang, Md. Islam, A. Kishore, P.A. Furth

Study supervision: P.A. Furth

This work was supported by grant NIH NCI RO1 CA112176 (to P.A. Furth), King Abdullah Scholarship Program, Ministry of Higher Education, Kingdom of Saudi Arabia (to S.J. Alothman), and NIH NCI 5P30CA051008 (Histology and Tissue, Genomics and Epigenomics, and Animal Shared Resources).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.