The PR status of the originating cell of ER/PR-negative mouse mammary tumors

Abstract

Progesterone receptor (PR) is usually co-localized with estrogen receptor (ER) in normal mammary cells. It is not known whether ER/PR-negative human breast cancer arises from an ER/PR-negative cell or from an ER/PR-positive cell that later lost ER/PR. Using intraductal injection of a lentivirus to deliver both an oncogene (ErbB2) and floxed GFP in PR^Cre/+mice, whose Cre gene is under the control of the PR promoter; we were able to trace the PR status of the infected cells as they progressed to cancer. We found that the resulting early lesions stained negative for PR in most of the cells and usually retained GFP. The resulting tumors lacked ER and PR, and 75% (15/20) of them retained the GFP signal in all tumor cells, suggesting PR was never expressed throughout the evolution of a majority of these tumors. In conclusion, our data demonstrate that ErbB2-initiated ER/PR-negative mammary tumors primarily originate from the subset of the mammary epithelium that is negative for PR and probably ER as well. These findings also provide an explanation for why antihormonal therapy fails to prevent ER-negative breast cancers.

INTRODUCTION

Approximately 10-15% of epithelial cells lining human mammary ductal tree produce estrogen receptor (ER) (1). These ER⁺ cells normally do not proliferate, but in response to estrogen they make paracrine factors that stimulate neighboring ER^- cells to undergo proliferation (1-4). Approximately 75% of human in situ and invasive breast carcinomas produce ER in at least 1% of their tumor cells (5). These ER⁺ cells proliferate, and these tumors as a whole usually depend on estrogen for survival (5). For this reason, breast cancers with at least 1% of ER⁺ cells are clinically designated ER⁺ tumors. Approximately 25% of in situ and invasive breast cancers have none or few ER⁺ cells (<1%) (6), and these ER^- cancers grow independent of estrogen signaling and require non-hormonal therapies.

The overwhelming majority of the clinically observed precancerous lesions of the breast produce ER in nearly all the cells (7, 8). For this reason, it is widely assumed that ER⁺ breast cancers have an origin in ER⁺ cells of the normal breast epithelium, which later gain the potential to proliferate and form ER⁺ precancerous lesions (7-9). For the same reason, it is also assumed that ER^- breast cancers arise from an ER⁺ precursor, which later lose ER in the evolution to cancer (8). However, it is also possible that ER^- tumors arise from one of the few ER^- cells within a generally ER⁺ precancerous lesion, which may originate from an ER⁺ or ER^- cell of the normal breast epithelium. In addition, it is conceivable that an ER^- cancer may arise from an ER^- cell of the mammary epithelium which then becomes an ER^- precancerous lesion that rapidly progresses to cancer and therefore is rarely detected in clinical samples. For example, although ErbB2 is amplified and overexpressed in 25% of human breast cancers, ErbB2 alterations have not been detected in atypical ductal hyperplasia(10-13). It is possible that ErbB2 amplification causes a rapid progression from a precancerous lesion such as atypical ductal hyperplasia (ADH) to ductal carcinoma in situ (DCIS) or invasive cancer, thus minimizing the chance of clinical detection of HER2-altered ADH.

To date, many genetically engineered mouse models have been generated to study mammary tumor evolution and progression (14-18); however, the genetic lesions in these models were introduced indiscriminately into both PR^- and PR⁺ cells. Thus, until now it has been difficult to use these models to investigate whether PR^- tumors actually originate from PR^- or PR⁺ cells or both. Here, we used a mouse model that allowed us to trace the PR status of the cell of origin of PR^- mammary tumors, and discovered evidence that most PR^- tumors initiated by ErB2 or PyMT have an origin in PR^- cells.

RESULTS AND DISCUSSION

Cre-mediated recombination is frequently used for tracing the cell of origin in development and cancer. In a mouse line with the Cre gene knocked into one ER gene locus and also containing a floxed reporter gene, all tumor cells within a cancer would undergo Cre-mediated recombination if the cancer arose from an ER⁺ cell, and no tumor cells within a cancer would undergo recombination if the cancer arose from an ER^- cell and ER was never expressed in the evolution to cancer either. However, a mouse line with Cre knocked into the ER locus has not been reported. PR is a classical ER transcriptional target which is co-expressed in 96% of ER⁺ mammary cells in adult mammary glands of the human and in similar high percentages in the mouse (1, 19). A mouse line with Cre knocked into the PR locus has been described (20). Therefore, in this study, we utilized PR^Cre/+ female mice (20) in combination with our previously reported intraductal injection of lentivirus (FUCGW) (14) to both induce tumors and report the PR status of the initially infected cells. The FUCGW vector was modified to express both an oncogene for tumor initiation and an IRES-driven GFP reporter that is flanked by the loxP recombination sites (Supplementary Figure 1A). Expression of PR in an infected cell would delete proviral GFP and cause this cell and all of its progeny to stop producing GFP (Supplementary Figure 1C&D). In contrast, if the cell destined to become PR^- mammary cancer does not express PR any time prior to full transformation, tumor cells in the resulting cancer would retain GFP (Supplementary Figure 1B&D). Therefore, GFP expression in the resulting PR^- tumors would demonstrate that the tumors arose from a PR^- cell, while a lack of GFP in the resulting tumors would reveal that PR was expressed some time in the transition from a normal epithelial cell to malignancy. Patchy GFP signal within a cancer would suggest that the tumor originated from a PR^- cell and that after the cancer has already formed, PR was switched on in some of the tumor cells.

First, we used the parental and GFP-expressing FUCGW (14) to confirm that this lentivirus has no preference to infect ER⁺/PR⁺ mammary epithelial cells or other cell subsets. We intraductally injected one set of mammary glands (L2, 3, 4) of 12-week-old PR^Cre/+ mice with FUCGW (IU:10⁸), and left the contralateral mammary glands (R2, 3, 4) un-injected as a baseline control. At 2 days post infection, we collected both the infected and un-infected mammary glands. We co-stained the infected mammary glands for both GFP and ER or both GFP and PR, so that we could determine the percentage of ER⁺ or PR⁺ cells within the GFP⁺ (and thus the infected) cell population. As shown in Figure 1A-C, 31.1±2.8% and 25.6±8.4% cells among the GFP+ infected mammary epithelial cells were positive for ER and PR respectively. These proportions of ER- or PR-positive cells among infected epithelial cell population are not significantly different from the percentages of ER⁺ or PR⁺ cells among non-infected mammary epithelial cells (28.0±7.9% for ER, p=0.48; 28.6±13.6% cells for PR, p=0.46), indicating that this viral vector does not have a preference for a particular hormonal status. In addition, we asked whether this lentivirus has a preference for proliferating cells. We co-stained the infected mammary glands for both GFP and Ki67, so that we could determine the percentage of Ki67⁺ cells within the GFP⁺ infected cell population. As shown in Figure 1D&E, 20.7±4.3% of the GFP⁺, infected mammary epithelial cells were positive for Ki67. This proliferation index is similar to what we have reported in normal ducts (17±12%) (21). Therefore, this lentiviral vector has no preference for proliferation status of the cell that it encounters.

Figure 1. The ER, PR status and proliferation rate of infected mammary epithelial cells.

A. Representative images of immunofluorescence staining of sections of mammary glands from mice infected by FUCGW (lenti-GFP). Left #2, 3, 4 mammary glands of 12-week old female PR^Cre/+ mice (n=5; on mixed strain background of FVB/N and C57BL6) were infected with lenti-GFP (10⁸IU) by intraductal nipple injection, while contralateral mammary glands (right #2, 3, 4) were used as the un-injected control. Two days after viral injection, the infected and un-infected mammary glands were collected, sectioned and co-stained for GFP and either ER or PR.

B-C. Distribution of ER⁺ (B) and PR⁺ (C) cells in normal mammary ducts and among lenti-GFP-infected mammary epithelial cells. At least 5 normal or infected ducts (n=5 mice per group) were photographed to quantify the percentages of ER⁺ or PR⁺ cells in uninfected, normal epithelial cells and in GFP⁺ epithelial cells. There was no significant difference between normal and infected ducts for each sample in B-C. D-E. Representative image (D) and quantification (E) of Ki67 staining in lenti-GFP-infected mammary epithelial cells. At least 5 infected ducts (n=5 mice per group) were photographed to quantify the percentage of Ki67⁺ cells in GFP⁺ infected epithelial cells. Our previously reported (21) quantification of Ki67 staining in normal mammary epithelial cells is shown here for comparison. PR^Cre/+ mice were randomized with wild type mice in the same cage. Five female mice were used to ensure detection of the adequate infected cells. Animal experimentation was approved by the Institutional Animal Care and Use Committee of Baylor College of Medicine, and conducted under the guidelines described in the NIH Guide for the Care and Use of Laboratory Animals. Lentiviral culture and mammary intraductal injection has been previously described (14). For immunofluorescence staining, the antibodies used were anti-GFP (1:100; JL-8; Clontech, Mountain View, CA, USA), anti-ER (1:400; MC-20; Santa Cruz Biotech; Dallas, TX, USA), anti-PR (1:100; A0098; Dako Denmark, Denmark), anti-Ki67 (1:200; NCL-Ki67p; Leica, Buffalo Grove, IL, USA). Quantification of all staining was done blindly. A two-tailed Student’s t test was applied for all comparisons, and p values of 0.05 or less were defined statistically significant. Continuous variable results were always reported as Mean ± SEM for each group.

To test whether this lentivirus could be modified to report PR status of infected cells during tumor evolution, we first inserted an IRES-GFP flanked by loxP sites downstream of the gene encoding PyMT in FUCGW to obtain lenti-PyMTfIG. PyMT activates Src and PI3K to initiate rapid mammary tumorigenesis, and we have reported that virus-mediated intraductal delivery of PyMT leads to oligoclonal tumorigenesis with a median latency of 12.5 days (21-25). We infected 15 PR^Cre/+ mice (age = 12 weeks) with lenti-PyMTfIG virus for early lesion and tumor studies. At five days post infection, we collected the infected mammary glands from five infected mice. As shown in Figure 2A, among the 50 early lesions examined by co-immunofluorescence for GFP and either PR or ER, all 13 PR⁺ lesions lacked GFP while all 26 GFP⁺ lesions lacked PR, indicating that GFP was indeed excised in PR⁺ cells during tumor initiation. Eleven early lesions were negative for both GFP and PR. These lesions likely originated from PR⁺ cells that later lost ER/PR expression as hormone receptor loss is known to occur during tumorigenesis due to aberrant growth factor signaling and as a result of expression of PyMT (26).

Figure 2. PR-negative PyMT tumors primarily arise from PR^- normal mammary epithelial cells.

A. Representative images showing three types of lenti-PyMTfIG-induced early lesions: GFP⁺/ER^-/PR^-, GFP^-/ER⁺/PR⁺, GFP^-/ER^-/PR^-. 12-week-old PR^Cre/+ mice (n=15) were infected with lenti-PyMTfIG virus (IU: 10⁵). Five mice were euthanized at five days post infection, and the remaining mice were euthanized when they developed a tumor at 1.0 cm in diameter.

B. Representative images showing that lenti-PyMTfIG-induced tumors lack PR and ER. Insets show PR or ER positive cells in adjacent normal ducts.

C. Bar-graph showing percentages of GFP⁺ cells in tumors arising in lenti-PyMTfIG-infected PR^Cre/+ mice.

We also monitored the remaining 10 infected mice for tumors. As expected, all mice developed tumors within 30 days with a medium latency of 9 days (data not shown). At approximately 1 cm in diameter, tumors were dissected for analysis. As expected, all 10 tumors were ER^-/PR^- (Figure 2B). As predicted, these presumably oligoclonal tumors harbored both GFP⁺ and GFP^- cells in the great majority of them (7/10), indicating that they had a mixed origin in both PR^- cells and PR⁺ cells, the latter of which excised GFP while also losing PR expression in the progression to full blown tumors. Three tumors lost GFP completely (#8, #9, &#10 in Figure 2C), indicating that all clones in these tumors originated from PR⁺ cells that later lost PR. However, we cannot exclude the possibility that some of the GFP-negative cells in these 10 tumors analyzed could originate from PR^- cells that transiently gained PR expression before losing it again in the eventual PR^- tumor. Together, we demonstrate that the floxed GFP in this lentivirus is indeed excised in response to Cre expression in PR^Cre/+ mice and that GFP expression is indicative of lack of PR expression anytime in clonal evolution while lack of GFP in a tumor may not indicate an origin in PR⁺ cells, as the approach was initially designed.

Having shown that this approach is suitable for studying the PR status of PR^- mammary tumors, we generated lenti-ErbB2fIG (ErbB2 plus floxed IRES GFP) to test the PR status of tumors initiated by ErbB2. We have previously reported that intraductal injection of lentivirus expressing a constitutively activated version of ErbB2 can induce mammary tumors (27). First, we confirmed that lenti-ErbB2fIG can generate early lesions within a few weeks and ER/PR-negative tumors later on. We intraductally injected 18 PR-wild type mice (age = 12 weeks), and euthanized 7 infected mice at two weeks post infection for early lesion confirmation and monitored the remaining 11 mice for tumor appearance. Early lesions were abundant at two weeks post viral injection, and 2.4±0.3% of cells in these lesions produced PR, while 3.7±0.9% stained for ER (Supplementary Figure 2A). Tumors formed with a median latency of 4 weeks (data not shown). Among 17 tumors collected from the 11 infected mice, PR⁺ or ER⁺ cells were rare (Supplementary Figure 2B). We also confirmed that GFP was indeed expressed in all infected cells in both early lesions and tumors by co-staining for the HA tag in ErbB2 and GFP (Supplementary Figure 3).

Next, we intraductally injected this lenti-ErbB2fIG virus into twenty-three 12-week-old PR^Cre/+ mice to study the PR status of the early lesions and tumors. At two weeks post infection, we euthanized 6 of the infected mice for early lesion studies. First, we validated that the floxed GFP in this provirus was responsive to Cre-mediated deletion in PR⁺ cells of PR^Cre/+ mice. As shown in Figure 3A&B, in the lesions arising in PR^Cre/+ mice, all GFP⁺ cells were negative for PR, and all PR⁺ cells were negative for GFP. Their mutually exclusive expression pattern validated that the floxed GFP DNA in our lenti-ErbB2fIG was responsive to Cre targeted to the PR locus and was indeed deleted in PR⁺ cells. As expected, PR was detected in some of the GFP⁺ cells in the lesions arising in PR-wild type mice (Figure 3A&B),

Figure 3. A majority of ErbB2-initiated early lesions originate from PR^- normal mammary epithelial cells.

A. Co-immunofluorescence showing GFP can be detected in PR⁺ cells of early lesions arising in wild-type mice, but not in PR⁺ cells of early lesions arising in PR^Cre/+ mice, demonstrating that the GFP gene cassette is indeed excised in PR-expressing cells in PR^Cre/+ mice. A photomicrograph of normal ducts was included to show the specificity of the staining for PR and GFP. Arrow indicates a cell positive for both GFP and PR. The mice (12 weeks of age; n=6) were intraductally injected with lenti-ErbB2fIG, and euthanized two weeks later for this assay.

B. Dot plot showing percentages of PR⁺ cells among GFP+ cells in early lesions from the above mice. Each dot represents one single lesion.

C. Photomicrographs showing two types of lesions (identified by staining for the HA-tag in ErbB2)—Lesion #1 retained GFP (lesion #1) while lesion #2 lost GFP (lesion #2). The mammary ducts are outlined by a yellow solid line, and lesions are outlined by a white dash line. The antibodies used were anti-HA (1:500; MMS-101P; Covance, Princeton, NJ, USA) and anti-GFP (1:200; FL; Santa Cruz Biotech, Dallas, TX, USA).

D. Bar-graph showing the percentages of GFP⁺ vs. GFP^- lesions in PR^Cre/+ mice (n=6 mice, 256 lesions total).

E. Bar-graph showing the percentages of GFP⁺ and GFP^- cells among the ErbB2 (HA)⁺ cells in early lesions.

As in the PR-wild type mice, early lesions arising in these PR^Cre/+ mice harbored a small percentage of ER⁺ and PR⁺ cells (approximately 4% and 3%, respectively; Supplementary Figure 2C). Among 265 lesions from 6 mice examined, 98.5% retained GFP (Figure 3C lesion#1; Figure 3D), while 1.5% of them lost GFP in at least 50% of the infected cells (Figure 3C lesion#2; Figure 3D). Likewise, in these 265 lesions, 97.4% of the HA+ cells were GFP⁺, while 2.6% were GFP^- (Figure 3E). Together, these observations indicate that majority of the ErbB2-initiated lesions originated from PR^- cells of the normal mammary epithelium.

Among the 17 PR^Cre/+ mice, 20 tumors were detected. All tumors harbored less than 1% of ER⁺ or PR⁺ cells (Figure 4B), and were designated ER- and PR-negative tumors. We stained these tumors for HA and GFP to determine if the tumors still retained GFP. Fifteen tumors still produced GFP (Figure 4A&B), indicating they originated from a PR^- cell of the mammary epithelium and never expressed PR in the evolution to malignancy. One tumor harbored approximately 30% GFP⁺ cells (#16 in Figure 4B), indicating that it originated from a PR-negative cell, but after the tumor had already formed, some of the cells within the tumor mass switched on PR expression. Alternatively, this tumor might be oligoclonal based on our previous study of ErbB2-induced tumors (5), and originated from both PR⁺ and PR^- cells. Four tumors lost GFP in all tumor cells examined (#17, #18, #19, & #20 in Figure 4B), indicating that they originated either from a PR⁺ cell or from a PR^- cell that later gained PR expression in its evolutionary path to a tumor. Together, these data demonstrate that ErbB2⁺/ER^-/PR^‐ tumors primarily arise from the PR^- subset of the mammary epithelium.

Figure 4. PR-negative ErbB2 tumors primarily arise from PR^- normal mammary epithelial cells.

A. Representative images of immunofluorescence staining of lenti-ErbB2fIG-induced tumors (n=20) from PR^Cre/+ mice (n=17).

B. Bar-graph showing percentages of ER⁺, PR⁺, and GFP⁺ cells in tumors arising in lenti-ErbB2fIG-infected PR^Cre/+ mice.

Collectively, our data demonstrate that ErbB2-initiated ER/PR-negative tumors primarily arise from the PR^- subset of the mammary epithelium. Based on our knowledge, this is the first report of tracing the hormonal status of mammary tumors. Our finding also has important clinical implications. It potentially explains why antihormonal therapy has been ineffective in preventing ER^- cancer—we hypothesize that upon the gain of genetic alterations such as ErbB2, ER^-/PR^- cells undergo estrogen-independent proliferation (21) and rapidly pass through the early lesion stages and become a clinical cancer. Therefore, to prevent ER^- breast cancers, therapeutic agents that target pathways that are unrelated to estrogen signaling but are important for proliferation or survival of the precursor of these ER^- cancers should be considered. Jak2-STAT5 signaling is activated in early lesions of ER^- breast cancer, intermittent blockade of Jak2-STAT5 signaling has been reported to prevent ER^- mammary tumors (27), and when combined with antihormonal therapy, may even prevent both ER^- and ER⁺ breast cancer.