A Multistage Murine Breast Cancer Model Reveals Long-Lived Premalignant Clones Refractory to Parity-Induced Protection

Epidemiology studies provide some of the best evidence that breast cancer is a multistage disease. Exposure to carcinogens, such as ionizing radiation, increases breast cancer risk, yet the interval between exposure and clinically detectable cancer typically spans decades (1, 2). Mechanistically, carcinogen-induced driver mutations have been proposed to create “initiated” premalignant clones, which are long-lived and predisposed to progress to frank malignancy. Alternatively, certain carcinogen exposures may not induce driver mutations directly, but instead may induce a protumorigenic microenvironment (e.g., chronic inflammation) that fosters the persistence and/or outgrowth of initiated clones (3, 4). However initiated clones originate, evolutionary modeling suggests that they frequently persist undetected for many years before finally manifesting clinically (5). Throughout this protracted latency period, initiated clones presumably elude detection by remaining small in size and/or generating only modest changes in histomorphology. Concordantly, analysis of carcinogen-exposed human skin and esophagus specimens by high throughput, multi-region DNA sequencing has revealed normal appearing epithelium to be a mosaic of mutant clones, many of which harbor mutations in cancer-associated genes (6, 7).

High-throughput sequencing of breast cancer genomes has generated detailed catalogs of driver mutations (8–10), but questions remain regarding the order in which these mutations accrue, as well as their stage-specific roles during tumor progression. The histopathology of the direct precursor breast lesions that ultimately beget invasive breast cancers remains poorly defined, which complicates efforts aimed at identifying, monitoring, or eliminating high-risk clones. Even when breast lesion histopathology correlates with increased cancer risk, it can remain unclear whether such lesions merely serve as risk markers or whether they comprise the definitive precursors (i.e., clonal antecedents) to subsequent breast cancers (11). Therefore, cancer models remain essential for interrogating mammary carcinogenesis across disease stages. Cell culture techniques and transplantable cell lines can be useful for modeling later stages of breast cancer progression, but autochthonous tumor models are essential for modeling earlier, subclinical disease stages.

From the very start, skin cancer modeling efforts provided an experimental framework for successful breast cancer modeling. Beginning in the 1930s, protocols for generating carcinogen-induced rodent skin cancers were adapted to generate the first transplantable mammary cancers (12). Most notably, carcinogen-induced rat mammary cancer models became widely used and proved crucial for elucidating the biology of hormone-dependent breast cancer (13). Later, rat mammary cancers provided the key preclinical model for developing antihormonal agents as therapeutics for breast cancer treatment and chemoprevention (14). Shortly after the first transforming oncogene mutations were identified, rodent skin and mammary tumors were found to share a key driver event, namely carcinogen-induced Hras mutation, highlighting a remarkable genetic convergence between these experimental models (15, 16). Whereas rats show a marked susceptibility to mammary tumors following carcinogen exposure, mice are relatively resistant. For example, a variety of rat strains reliably develop mammary tumors after a single carcinogen exposure administered during puberty (17–19). In contrast, mice reliably develop mammary cancers only when carcinogen exposures are administered repeatedly or supplemented by stimulation with exogenous hormones (20, 21). The biological explanation for this species difference remains unknown.

Transgenic mouse models reinforce the concept that mammary carcinogenesis is a multistage process. For example, whereas Wnt pathway activation drives synchronous onset of diffuse mammary hyperplasia in a matter of days, Wnt-driven mammary cancers instead arise stochastically over a period of months (22, 23). Unlike Wnt-driven mammary hyperplasia, most Wnt-driven cancers acquire cooperating somatic mutations that create oncogenic Hras alleles (24–26), implicating Hras activation as a rate-limiting step in carcinogenesis. Supporting this model, we showed previously that iWnt transgenic mice (engineered for inducible, mammary-specific Wnt pathway activation), when subjected to Wnt induction followed by a single subcarcinogenic 7,12-dimethylbenzanthracene (DMBA) exposure (i.e., a DMBA dose insufficient to generate tumors in wild-type mice), rapidly develop multiple mammary cancers, nearly all of which bear stereotyped Hras^Q61L mutations attributable to DMBA (27). Thus, the synergy between DMBA exposure and Wnt-driven mammary hyperplasia during mammary carcinogenesis closely resembles synergy between DMBA and the regenerative skin hyperplasia induced by repeated topical application of 12-O-tetradecanoyl-phorbol-13-acetate (TPA) in classic multistage skin carcinogenesis. Strikingly, in both models, chronic induction of hyperplasia culminates in the outgrowth of DMBA-initiated, Hras^Q61L-mutant epithelial cell clones.

Given these deep parallels between DMBA-initiated skin and mammary cancers, we set out to develop a multistage DMBA/iWnt mammary cancer model. However, at the outset, we recognized that DMBA/iWnt mammary tumorigenesis differs from DMBA/TPA skin carcinogenesis in at least two key ways. First, in classic multistage skin cancer modeling, promotion alone is insufficient for generating tumors (28). In contrast, strictly speaking, Wnt pathway activation can act as a “complete carcinogen” in the mammary gland, because iWnt mice monitored over many months of chronic Wnt induction eventually develop solitary mammary tumors (22, 29). Second, whereas the first skin tumors detected during classic multistage skin cancer modeling are benign papillomas, the first mammary tumors detected in DMBA/iWnt experiments are invasive cancers, as discussed below. In light of these differences, when adapting our DMBA/iWnt protocol for multistage breast cancer modeling, we used classic terminology by referring to the DMBA “initiation” step, but adopted modified terminology by referring to a Wnt-driven “progression” (rather than “promotion”) step.

Mice were bred and maintained at the Pennsylvania State University College of Medicine pathogen-free rodent facility with free access to water and chow. All experimental protocols were approved by the Pennsylvania State University College of Medicine's Institutional Animal Care and Use Committee. The MMTV-rtTA and TetO-Wnt1 transgenic mouse lines (refs. 22, 30; gifts from Dr. Lewis Chodosh, University of Pennsylvania) were generated and maintained on an inbred FVB/N background; the TetO-H2B-GFP transgenic line (31) was obtained from the Jackson Laboratory (Stock #005104), then backcrossed to FVB/N more than 20 generations prior to generating offspring for this study. Mice were genotyped using PCR-based assays performed on tail snip-derived genomic DNA. PCR primer pairs included the following. MMTV-rtTA transgene: 5′-ATCCGCACCCTTGATGACTCCG-3′ with 5′-GGCTATCAACCAACACACTGCCAC-3′ TetO-Wnt1 transgene: 5′-TGCGGTTCCTGATGTATTTTGC-3′ with 5′-TGCATTCCTTTGGCGAGAGG-3′. TetO-H2B-eGFP transgene: 5′-AAGTTCATCTGCACCACCG-3′ with 5′-TCCTTGAAGAAGATGGTGCG-3′.

For doxycycline treatment, standard mouse chow was replaced with medicated chow containing 2 g/Kg doxycycline (Bio-Serv). For chemical carcinogenesis, DMBA (Sigma-Aldrich D3254) was dissolved in sesame oil (Sigma-Aldrich N3385) at 5 mg/mL. Mice were administered a 1 mg DMBA dose by oral gavage at the indicated time points.

Mice were examined daily for tumors by palpation. For detailed serial monitoring of tumors, caliper measurements and body photographs were obtained while mice were transiently anesthetized using inhaled isoflurane. Mice were euthanized and subjected to necropsy and tumor harvest when tumor burden reached predetermined endpoints described in the main text. Tumors were subdivided and either fixed with 4% paraformaldehyde (PFA) for 2 hours prior to histology processing or snap-frozen on dry ice prior to storage at −80°C.

Genomic DNA was isolated from frozen tumor tissue using Promega Maxwell 16 Tissue DNA Purification Kit (Promega AS1030). A relevant Hras gene segment was PCR-amplified using the following primers: 5′-GGGTCAGGCATCTATTAGCCGTC-3′ with 5′-GCCGAGACTCAACAGTGCGAG. PCR products were purified using either the Qiaquick PCR purification kit (28104) or Affimetrix ExoSAP-IT (75001) kit. Sanger sequencing was performed by Genewiz LLC, and results were analyzed using SnapGene software.

For carmine staining, mammary glands were spread on glass slides, fixed in 4% PFA for 2 hours, rinsed in 70% ethanol, rehydrated, and stained with 0.2% carmine alum for at least 24 hours. After dehydration in graded solutions of ethanol (70%, 80%, 90%, and 100%), glands were cleared with methyl salicylate, then photographed using a Nikon SMZ1000 Zoom stereomicroscope. For fluorescent imaging of mammary gland whole-mounts, glands were spread on glass slides, fixed in 4% PFA for 15 minutes, rinsed with PBS, and then photographed using a Leica MZ FLIII Fluorescence Stereomicroscope with a standard GFP emission filter. Image reconstructions were generated using ImageJ. Subclinical lesions and neighboring glandular tissues were identified by microscopy and were excised and collected on the stage of the dissecting microscope.

Fragments of DMBA/iWnt primary tumors were implanted onto the flanks of syngeneic female FVB host mice maintained on chronic doxycycline treatment. When explanted tumor outgrowths reached volumes of 300–500 mm³, the host mice were either left untreated or subjected to ovariectomy as indicated. Successful ovariectomy was confirmed both by histologic assessment of the extirpated tissue and by documenting atresia of the uterus at subsequent necropsy. Tumor volumes were calculated by the modified ellipsoidal formula: tumor volume = 0.5 × (tumor length) × (tumor width)².

Tumor-free survival curves were compared across cohorts using the log-rank test. In separate analyses, time-to-first tumor and tumor multiplicity were compared across cohorts using one-way ANOVA with the Tukey correction for multiple comparisons. Datasets were not evaluated for normality. Cohort sizes were not predetermined using a power calculation, which in principle poses an increased risk of type II error. Instead, because high tumor multiplicity rendered each mouse unusually informative, experiments were performed using cohort sizes more typical of pilot studies (4–6 mice/cohort). Most comparisons were made among contemporaneous cohorts. To minimize mouse usage, some comparisons employed a cohort from a prior experiment as an “historical control.” Comparisons involving historical controls, whenever they are employed, are indicated as such in the Results section. All graphs and statistical analyses were generated using GraphPad Prism 8.

With the goal of developing a DMBA/iWnt multistage mammary carcinogenesis model, we confirmed that no mammary tumors arose in a control cohort of iWnt mice subjected to DMBA exposure alone (“I-only” for initiation-only), indicating that the 1 mg dose administered by gavage is subcarcinogenic. In a second control cohort, Wnt pathway activation in DMBA-naïve iWnt mice (“WP-only” for Wnt progression-only) yielded solitary, late-onset mammary tumors after a mean of 20 weeks of doxycycline treatment (Fig. 1A), consistent with previous studies (22, 29). In our published DMBA/iWnt work, DMBA exposure always was administered 1 week after doxycycline (Dox)-induced Wnt pathway activation (27). To test whether DMBA's profound action as a cocarcinogen requires the presence of preexisting Wnt-driven mammary hyperplasia, we reversed the experimental sequence by subjecting 5-week-old, doxycycline-naïve iWnt female mice to DMBA exposure up-front, then activating chronic Wnt-driven mammary hyperplasia 4 days later. This I-4d-WP protocol (initiation-4 day interval-WP) better recapitulates the experimental sequence in classic multistage skin carcinogenesis, where DMBA-mediated initiation precedes TPA-mediated promotion. iWnt mice subjected to the I-4d-WP protocol rapidly developed multiple palpable mammary tumors within 3 weeks of Wnt pathway activation (21 ± 1.1 days; Fig. 1A–C). Compared with the solitary mammary tumors that arose in the WP-only cohort (0.80 ± 0.06 mammary tumors/mouse), tumor multiplicity increased more than 20-fold in mice from the I-4d-WP cohort (18.6 ± 1.8) when examined at necropsy 5–6 weeks after Wnt activation (Fig. 1D). We infer that a one-time DMBA exposure, although subcarcinogenic on its own, reliably generated numerous initiated mammary epithelial cells (iMEC) in each mouse, which strictly require Wnt-driven progression to transform into malignant tumors.

Mammary tumors generated in the context of DMBA exposure were histologically indistinguishable from DMBA-naïve iWnt tumors. Specifically, tumors from the I-4d-WP cohort were classic P-type adenocarcinomas (32), which showed mixed lineage differentiation and well-developed stroma, frequently accompanied by small foci of squamous differentiation. Moreover, mammary tumors generated by the I-4d-WP protocol were invasive, because explanted tumor fragments readily produced tumor outgrowths when implanted onto the flanks of doxycycline-treated, syngeneic female host mice (Supplementary Fig. S1).

A canonical feature of classic multistage skin cancer modeling is the irreversibility of the initiation step, that is, initiated skin cells persist in a clinically silent state, yet remain available for, and amenable to, TPA-mediated promotion months later (33, 34). To determine whether transformation-competent iMECs are similarly long lived, we tested modified DMBA/iWnt protocols in which the 4 day time interval separating initiation and progression (I-4d-WP) was increased to either 4 weeks (I-4wk-WP) or 12 weeks (I-12wk-WP). Remarkably, neither delay in Wnt-driven progression resulted in iMEC dropout. Instead, mice subjected to delayed Wnt activation remained poised for rapid outgrowth of invasive tumors following Wnt-driven progression (Fig. 1A and B). In fact, palpable mammary tumors arose even faster in the context of delayed Wnt activation (15.3 ± 1.0 days for the I-4wk-WP cohort; 13.8 ± 1.4 days for the I-12wk-WP cohort), suggesting iMEC clones may undergo modest subclinical expansion in the weeks preceding Wnt-driven progression. Importantly, DMBA/iWnt mice subjected to delayed Wnt activation remained comparably predisposed toward high tumor multiplicity (17.8 ± 2.4 tumors/mouse for the I-4wk-WP cohort; 16.5 ± 0.29 tumors/mouse for the I-12wk-WP cohort; Fig. 1D), indicating that most, if not all DMBA-induced iMECs are long lived. In addition, tumors never arose in DMBA-exposed mice during even extended intervals preceding Wnt activation, consistent with a strict requirement for Wnt activation during iMEC outgrowth.

In multistage skin carcinogenesis, initiation involves DMBA-induced Hras^Q61L mutations that generate long-lived premalignant clones. To confirm that DMBA induces the identical Hras^Q61L mutation in long-lived iMECs, Sanger sequencing was performed on PCR-amplified Hras alleles derived from mammary tumors that arose with and without prior DMBA exposure. Wnt-driven tumors from DMBA-naïve mice (WP-only condition) frequently harbored cooperating Hras mutations (22/34 tumors, 65%), which were distributed among three distinct recurring mutations (G12E, Q61R, and Q61L; Fig. 1E), consistent with previous reports (24–26). In contrast, nearly all Wnt-driven tumors from DMBA-exposed mice harbored an activating Hras mutation (95% or more of tumors analyzed, at least 41 tumors analyzed per condition), which nearly always was the Hras^Q61L mutation stereotypically induced by DMBA exposure in skin cancer modeling. This overwhelming predilection for acquiring an Hras^Q61L mutation was observed regardless of whether Wnt-driven progression occurred shortly after DMBA exposure (I-4d-WP condition) or occurred weeks later (I-4wk-WP and I-12wk-WP conditions; Fig. 1E), implicating Hras^Q61L mutations in the genesis of long-lived iMECs.

Next, we determined when during their genesis and outgrowth iMECs produce detectable histopathology by performing microscopic examination of carmine-stained mammary gland whole-mounts and hematoxylin and eosin (H&E)-stained tissue sections. Absent activation of Wnt-driven progression, iMECs produced no overt mammary lesions, because glands harvested from mice either 4 weeks or 12 weeks after DMBA exposure were indistinguishable from those of age-matched, DMBA-naïve controls. Although iMECs generated using the I-4d-WP protocol reliably generated a dozen or more palpable mammary tumors within 3–4 weeks of WP, time course analyses of carmine-stained mammary gland whole-mounts showed that the onset of corresponding microscopic mammary ductal lesions following Wnt activation was surprisingly slow. Mammary glands from iWnt mice subjected to the I-4d-WP protocol harvested at day 6 of WP lacked any focal lesions discernible against the background Wnt-driven hyperplasia induced throughout the entire gland (0 focal lesions identified in 10 glands, 0%). Even out to day 14 of WP, most glands remained free of focal mammary ductal lesions, and the few lesions beginning to emerge were small (four microscopic focal lesions identified in 10 glands, 40%; Fig. 2A). Nonetheless, thereafter these microscopic foci grew quickly and synchronously such that just 5 days later (day 19 WP), all mammary glands harbored one or more obvious macroscopic lesions (10/10 glands, 100%; Fig. 2), the largest of which already were palpable in live mice prior to euthanasia and tissue harvest. Thus, iMECs comprise a cryptic reservoir of premalignant cells with an impressive capacity for rapid, Wnt-driven clonal expansion.

If DMBA-induced Hras^Q61L mutations mark the clonal link between iMECs and overt mammary cancers, then the subclinical mammary lesions that emerge during WP ought to bear Hras^Q61L mutations as well. To aid in the identification of subclinical lesions within freshly harvested tissue, a reporter gene was introduced into the iWnt model via crosses with a TetO-H2B-GFP transgenic line (refs. 31, 35; Fig. 2B). As expected, doxycycline-induced H2B-GFP expression in tri-transgenic offspring (MMTV-rtTA/TetO-Wnt1/TetO-H2B-GFP; hereafter iWnt/iGFP mice) permitted visualization of mammary epithelium by fluorescent microscopy (Fig. 2C). When mammary glands from iWnt/iGFP mice subjected to the I-4d-WP protocol were harvested after timed periods of Wnt-driven progression, day 11 WP glands lacked focal lesions (0/10 glands examined by fluorescence microscopy, 0%), and the first focal lesions emerged at day 14 WP (one focal lesion in 10 glands, 10%). In contrast, every mammary glands harvested at day 18 WP harbored one or more focal mammary lesions (40/40 glands, 100%; Fig. 2C), some of which were macroscopic. Subclinical focal lesions from these glands were excised on the stage of a dissecting microscope, along with separate samples of neighboring glandular tissue comprised of diffuse Wnt-driven mammary hyperplasia. Sanger sequencing performed on PCR-amplified Hras alleles showed that genomic DNA derived from focal lesions invariably carried Hras^Q61L mutations, whereas neighboring epithelium lacked detectable Hras mutations (Fig. 2C). In sum, iMECs subjected to Wnt activation yielded focal mammary lesions that grew into overt mammary cancers in a quasi-synchronous manner, raising the possibility that the Wnt-driven progression from iMEC to frank malignancy may be a single-step transition. In this way, Wnt-driven progression of iMECs differs from TPA-mediated promotion of initiated skin cells, which instead yields papillomas in a stochastic manner, and only after many months.

Both epidemiology studies (1, 2) and rat models (19, 36) suggest that more mature stages of breast development may show decreased carcinogen susceptibility. In our initial DMBA/iWnt experiments, DMBA exposure was administered at mid-puberty to 5-week-old females, which are in the midst of end bud-mediated mammary ductal elongation. To test whether mature iWnt mammary glands show decreased susceptibility to DMBA-induced iMECs, we modified the I-4d-WP protocol by administering DMBA to adult 12-week-old iWnt females, in whom mammary ductal elongation is complete and terminal end buds have regressed. Despite mammary gland maturation, adult females remained susceptible to DMBA-induced iMECs, which again rapidly yielded multiple palpable mammary tumor outgrowths within 3 weeks of subsequent Wnt-driven progression (Fig. 3A). Similar to the iMECs acquired during puberty, iMECs acquired via adult DMBA exposure were long lived, because no iMEC dropout was observed upon increasing the interval separating initiation and progression to either 4 weeks (I-4wk-WP) or 12 weeks (I-12wk-WP). On the contrary, in the context of adult DMBA exposure, delayed WP trended toward producing more rapid tumor outgrowth and higher tumor multiplicity (Fig. 3B and C). As before, tumors arising from adult DMBA-exposed mice nearly always harbored the stereotyped DMBA-induced Hras^Q61L mutation (Fig. 3D).

Next, we examined whether parity-induced protection schemes reduce tumor incidence in our multistage breast cancer model. While established Wnt-driven mammary tumors have been shown to grow in a hormone-independent manner (37), we reasoned that this hormone independence might not be shared by antecedent iMEC clones, which grow slowly if at all prior to Wnt activation. First, we tested for parity-induced protection using a preexposure prophylaxis protocol (Fig. 4A), wherein iWnt mice began a full-term pregnancy at 5 weeks of age, nursed pups for 1 week up until forced weaning, then underwent mammary gland involution prior to adult DMBA exposure. Because the length of the involution period optimal for parity-induced protection is unclear, we tested separate parity schemes incorporating either a 3-week involution period (P3wk scheme) or an 8-week involution period (P8wk scheme). Neither scheme conferred parity-induced protection, because parous iWnt females remained susceptible to DMBA-induced iMECs, which rapidly yielded multiple palpable mammary tumor outgrowths within 3 weeks of Wnt-driven progression (Fig. 4B and C), nearly all of which harbored the stereotyped DMBA-induced Hras^Q61L mutation (Fig. 4D).

Finally, we tested for parity-induced protection using a postexposure prophylaxis protocol, wherein iWnt mice were subjected to DMBA exposure at 5 weeks of age, then completed either the P3wk or P8wk parity scheme prior to beginning Wnt-driven progression (Fig. 5A). Once again, neither scheme conferred parity-induced protection. DMBA-induced iMECs persisted through postexposure parity and rapidly yielded multiple palpable mammary tumor outgrowths within 3 weeks of Wnt-driven progression (Fig. 5B and C). As before, nearly all of these tumors harbored the stereotyped DMBA-induced Hras^Q61L mutation (Fig. 5D).

Our DMBA/iWnt multistage mammary cancer model recapitulates canonical features of classic DMBA/TPA skin cancer modeling, albeit with several key differences. While DMBA is applied topically in skin models versus systemically (via gavage) in our mammary model, both modes of exposure induce stereotyped Hras^Q61L mutations, thereby generating initiated epithelial cells. In both skin and mammary tissue, initiation occurs irreversibly because both TPA-mediated promotion and Wnt-driven progression yield tumors efficiently, even when started months after completing the initiation stage. Interestingly, just as most DMBA/TPA skin papillomas regress when daily TPA application is withdrawn (38), iWnt tumors consistently regress when challenged with Wnt withdrawal (22, 26, 29), including DMBA/iWnt tumors. Despite these striking similarities, the models differ starkly in how initiated cells are provoked to yield tumors. Topical TPA meets the classic definition of a tumor promoter, because TPA-induced regenerative hyperplasia never induces skin tumors in the absence of initiation. In contrast, whereas Wnt1 transgene-driven mammary hyperplasia involves diffuse, benign overgrowth of mammary epithelium in the short term, Wnt1 acts as a so-called “complete carcinogen” over the long term, in that DMBA-naïve iWnt mice stochastically develop mammary carcinomas after months of Wnt activation.

Moreover, the DMBA/TPA and DMBA/iWnt protocols differ in that only DMBA/TPA provides direct experimental access to a clinically discernible benign lesion. The DMBA/TPA protocol produces benign papillomas that arise in a stochastic manner after several months of TPA application, and only a minority of these papillomas progress to squamous cell cancer during additional weeks of TPA (38). In contrast, the DMBA/iWnt protocol produces prompt, quasi-synchronous outgrowth of invasive, transplantable mammary cancers within 3–4 weeks of Wnt pathway activation. Therefore, DMBA-induced iMEC clones seemingly undergo a single-step malignant conversion. In light of these key differences, we refer to the Wnt-driven “progression” of iMECs, to distinguish this process from the TPA-driven “promotion” of initiated skin cells.

Unlike some epidemiology and rat chemical carcinogenesis studies, mature mouse mammary glands were no less susceptible to DMBA-initiated mammary tumorigenesis in our multistage modeling. This discrepancy may reflect species differences, but also may reflect differences in the type and/or level of carcinogen exposure employed. Notably, our modeling produced very high tumor multiplicity, suggesting that our DMBA/iWnt protocols may overwhelm protective mechanisms, which normally might restrict tumorigenesis either by preventing iMEC generation (e.g., via DNA repair) or enabling iMEC elimination (e.g., via immune surveillance). In future studies, it will be important to test revised protocols that employ clinically relevant breast carcinogen exposures (e.g., ionizing radiation) and/or produce clinically relevant tumor multiplicity (i.e., solitary tumors).

In multistage skin carcinogenesis, initiated skin cells persist life-long, leading some to conclude that the initiating Hras^Q61L mutation likely targets a stem cell. By analogy, DMBA may create iMECs by introducing mutations within a long-lived mammary stem cell compartment. In rats, the terminal end buds present during puberty have been proposed to harbor mammary cells that are uniquely susceptible to carcinogen-induced transformation (17). In the DMBA/iWnt model, terminal end buds were dispensable for the efficient induction of iMECs. Adult iWnt mice acquired DMBA-induced iMECs, even though ductal morphogenesis had been completed and terminal end buds had regressed. That said, in a previous study (26), we found that Hras-mutant subclones from iWnt mammary tumors have basal features, which may reflect their origin from within the basal MEC compartment, where mammary stem cells are believed to reside. Alternatively, the Hras^Q61L mutation, when targeted to a short-lived mammary progenitor cell, may render that cell long-lived. In lineage tracing studies performed in mouse intestinal epithelium, introducing an activating Kras mutation increased the likelihood that a cell would persist as a long-lived clone (39, 40).

Several features of our multistage breast cancer model complicate a straightforward translation of our findings to clinical scenarios. While activating Hras mutations are a recurring genetic event in breast cancer, they occur only rarely (<1% of cases; ref. 8), whereas such mutations are relatively common in skin cancer (∼20% of cases; ref. 41). That said, Hras mutations are key driver events in rodent breast cancer models with proven clinical utility. Although chemical carcinogen–induced rat mammary cancers carry highly recurrent Hras mutations, they nonetheless provided an indispensable preclinical model for the development of key antihormonal agents, including tamoxifen and aromatase inhibitors (14). Likewise, mutation-based activation of the Wnt pathway seems to be uncommon in breast cancer. With respect to modeling of parity-induced protection, some epidemiology data suggests that early parity primarily reduces the risk of hormone-dependent breast cancer (42). Consistent with this idea, established Wnt-driven mammary cancers have been shown to grow in a hormone-independent manner (37), which may help to explain why Wnt-responsive iMECs are refractory to parity protection schemes. Conversely, preventive antihormonal treatment with tamoxifen delays onset of mammary tumors in the classic MMTV-Wnt1 model, suggesting that different disease stages may show distinct hormonal requirements (43).

Thus far, iMECs remain abstract entities, whose existence is inferred from the requirement that DMBA/iWnt mammary cancers arise from clonal antecedents. Our early efforts to visualize latent iMEC clones in situ have proved unsuccessful thus far, leaving key features of iMEC biology uncharacterized. Outstanding questions include, whether latent iMEC clones are maintained as single cells or multicellular structures, and whether such cells remain quiescent or undergo dynamic turnover. Addressing these questions, among others, will require new experimental strategies. In the case of multistage skin carcinogenesis, numerous insights have emerged by superimposing lineage tracing methodology on the classic DMBA/TPA protocol (44). Similarly, lineage tracing performed in the context of DMBA/iWNT experiments might prove informative. In the long term, identifying the culprit iMECs that bridge carcinogen exposure and tumor progression should inform efforts to design rational breast cancer chemoprevention strategies. In the short term, ongoing work in our group is testing pharmacologic strategies aimed at eradicating iMECs using FDA-approved drugs.

No potential conflicts of interest were disclosed.

Conception and design: S. Li, E.J. Gunther

Development of methodology: S. Li, S.A. Gestl, E.J. Gunther

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Li, S.A. Gestl

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Li, S.A. Gestl, E.J. Gunther

Writing, review, and/or revision of the manuscript: S. Li, S.A. Gestl, E.J. Gunther

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Li, S.A. Gestl

Study supervision: E.J. Gunther

We thank members of the Gunther laboratory for critical review of the article. This work was supported by grants from the NCI to E.J. Gunther (R01CA152222 and R01CA212584) and funding received from the benefactors of the Jake Gittlen Laboratories for Cancer Research. Animal housing was provided through a facility constructed with support from a Research Facilities Improvement Grant (C06 RR-15428-01) from the National Center for Research 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.