Generation of the MIC construct and rtTA/MIC bigenic strain

The MIC construct was created using the pTE-mElf5-IRES-eGFP vector (a generous gift from Dr. C. Ormandy). Briefly, after removal of the mElf5 and eGFP transgenes, PyV mT cDNA was sub-cloned between the Tet-operator (TetO) and internal ribosome entry sequence (IRES), followed by sub-cloning of Cre recombinase cDNA downstream of the IRES (TetO-PyV mT-IRES-Cre recombinase; abbreviated as MIC) (Additional file 1: Figure S1). Derivation of the MIC strain was conducted in the Transgenic Core Facility in the Goodman Cancer Research Centre using standard pronuclear injection of FVB/N single cell embryos [11]. Progeny were screened for germ-line transmission of the MIC transgene by PCR genotyping. MMTV-reverse tetracycline transactivator (rtTA) transgenic mice were generated in the laboratory of Dr. Lewis Chodosh as previously described [9]. The ROSA26 Cre recombinase-activated β-galactosidase reporter strain (GTRosa) was generated in the laboratory of Dr. Phillipe Soriano as previously described [12]. All mice were housed in the animal facility at the Goodman Cancer Research Centre. Ethical approval was obtained for the use of animals and all experiments were done in accordance with the animal care guidelines at the Animal Resource Centre of McGill University. All transgenic animals described in this study were on a pure FVB/N background.

Doxycycline induction of rtTA/MIC mice and necropsy procedures

Experimental and control animals of at least eight weeks of age were given drinking water containing 2 mg/mL doxycycline (Sigma Aldrich, St. Louis, MO, USA) in light-blocking bottles each week. Mammary tumours were detected by physical palpation and animals were sacrificed at a total tumour burden of six cubic centimetres unless otherwise specified. Material from necropsied mice was flash frozen in liquid nitrogen (in some cases, tissues were set in an OCT medium before freezing) or fixed in 10% neutral buffered formalin (Leica Microsystems Inc., Concord, ON, Canada) and embedded in paraffin wax. Fixed and embedded mammary tissues were sectioned at 4 μm and either stained by haematoxylin and eosin (H&E) or processed further as indicated. H&E-stained lung sections (4 μm) were scanned with a Scanscope XT Digital Slide Scanner and for each sample the area occupied by metastases relative to the total lung area was quantified using Imagescope software (Aperio, Vista, CA, USA). The number and type of lesions in mammary gland sections were counted similarly by analysing 10 fields within a scanned image of each sample. Inguinal mammary glands were whole mounted on glass slides and processed for H&E-staining as described [6]. For resection surgeries, mice were anesthetised prior to excising a small piece of tumour for embedding and flash freezing; the incision was sutured and painkillers administered for one week.

Immunoblotting

Flash frozen mammary tissues were lysed as in [6] and 20 μg of protein were separated by SDS-PAGE. Membranes were immunoblotted with the following primary antibodies: E-cadherin (BD Biosciences, Mississauga, ON, Canada; 610182, 1:1,000); Cre recombinase (Novagen/EMD Millipore, Billerica, MA, USA; 69050, 1:1,000); P-EGFR (Y1068) (Cell Signaling/New England Biolabs, Pickering, ON, Canada; #3777, 1:1,000); EGFR (Cell Signaling #2322, 1:1,000); P-ErbB2 (Y1248) (Santa Cruz Biotechnology, Santa Cruz, CA, USA; #12352, 1:500); ErbB2 (Santa Cruz Biotechnology #284, 1:1,000); Hsp90 (Cell Signaling #4874); PyV mT (a generous gift from Dr. S. Dilworth, Ab750, 1:1000), c-Myc (Santa Cruz Biotechnology #764); P-PDGFRβ (Y1021) (Cell Signaling #2227, 1:1,000); PDGFRβ (Cell Signaling #3175, 1:1,000); α-tubulin (Cell Signaling #2125). Horse radish peroxidase (HRP)-conjugated secondary antibodies were purchased from Jackson ImmunoResearch Laboratories, West Grove, PA, USA. GE Amersham (Bai d'Urfe, QC, Canada) enhanced chemiluminescence detection reagents were used to visualize the immunoblots.

Immunohistochemistry

Sections of paraffin-embedded samples were immunostained as described previously [6]. Antigen retrieval was performed in 10 mM sodium citrate buffer. Sections were blocked in a solution of Power Block Universal Blocking Agent (Biogenex , Fremont, CA, USA) and incubated with primary antibodies: PyV mT (a generous gift from Dr. S. Dilworth, Ab762, 1:100) or Cre recombinase (Covance, Denver, PA, USA; PRB 106C, 1:600). In the case of PyV mT, the sections were incubated with anti-mouse secondary antibody conjugated to HRP (Dako, Burlington, ON, Canada; #K4006). For Cre recombinase, the sections were incubated first with biotinylated anti-rabbit secondary antibody (Jackson ImmunoResearch Laboratories) and then with avidin conjugated to HRP (Vector Labs, Burlington, ON, Canada; Vectastain Elite ABC kit #PK-6100). Development was carried out by exposing with DAB reagent (Dako #K3467). Slides were counterstained with 20% haematoxylin before mounting.

β-galactosidase assay

OCT-embedded sections of mammary tumours were processed and stained with X-gal as described previously [6].

qRT-PCR

Total RNA was extracted from flash frozen mammary tumours and lung lesions using a Qiagen AllPrep DNA/RNA Mini Kit (Qiagen Inc., Toronto, ON, Canada; #80204). cDNA was prepared by reverse transcribing the isolated RNA using M-Mulv Reverse Transcriptase (#M0253S), Oligo-dT(23VN) (#S1327S) and a murine RNase inhibitor (#M0314S) (all purchased from New England Biolabs). Real-time quantitative PCR was performed on the cDNA using LightCycler 480 SYBR Green I Master (Roche, Missisauga, ON, Canada; #04707516001) and run on a Roche LightCycler 480 instrument. Primers used for PyV mT were as follows: forward - 5’CCCGATGACAGCATATCCCC-3’; reverse - 5’CTTGTTCCCCCGGTAGGATC-3’. PyV mT transcript levels were normalized to GAPDH transcript levels.

DNA sequencing

Phospho-RTK array

The Proteome Profiler Mouse Phospho-receptor tyrosine kinase (RTK) Array Kit from R&D Systems/Cedarlane Corp., Burlington, ON, Canada (#ARY014) was used according to the manufacturer’s instructions. Flash frozen mammary tissues were lysed as in [6] and 200 μg was used for each array. Arrays spotted with RTK probes were incubated with biotin-conjugated phospho-tyrosine antibody (EMD Millipore #16-103, 1:1,000) followed by fluorescently-conjugated streptavidin (Mandel Scientific, Guelph, ON, Canada; #LIC-926-32230, 1:1,000) to allow for visualization on the Odyssey Imaging System (LI-COR Biosciences, Lincoln, NE, USA). Fluorescence intensities for each probe were normalized to a PBS probe (negative control).

Induction of MIC transgene expression in the mammary gland results in rapid tumour onset

Polyomavirus middle T antigen (PyV mT) and Cre recombinase cDNAs were sub-cloned into a pTE vector containing an internal ribosome entry sequence (IRES) to produce a TetO-PyV MIC transgene (Additional file 1: Figure S1). MIC virgin females were aged to one year without issue and the transgene did not disrupt normal breeding in either sex or nursing by females. MIC founder lines were crossed to the MMTV-rtTA strain to drive doxycycline-inducible transgene expression to the mammary epithelium [9]. Tumour onset in the original constitutive MMTV-PyV mT model occurs with relatively short onset, with virgin females developing mammary masses with a T₅₀ of 40 days of age [3, 6]. To evaluate whether this was also the case for the MIC model, we induced cohorts of rtTA/MIC, rtTA and MIC virgin female mice between 8 and 16 weeks of age with 2 mg/mL doxycycline which has been previously shown to lead to robust expression of the Tet-inducible transgene in an MMTV-rtTA background [9]. The minimum age of eight weeks for induction was chosen to ensure that the mammary epithelium would be almost fully developed. Induced animals were initially examined every other day by physical palpation for mammary tumour formation, alongside un-induced controls of the same genotypes. A single founder line in which mammary tumours were detected was chosen for further breeding to generate cohorts that would be more extensively characterized.

Mammary gland masses were detected in rtTA/MIC mice as early as four days post-induction, with 74.4% (29/39) of the cohort developing multifocal tumours within 16 days of induction (Figure 1A). The few animals that did not palpate within this short 16-day window of induction could be subdivided into two groups: those that developed tumours between 17 and 365 days post-induction (12.8%; 5/39) and those that remained tumour-free after one year of induction (12.8%; 5/39). Considering the entire rtTA/MIC induction cohort, the average tumour onset was 22.0 ± 7.1 days post-induction while the T₅₀ was 7 days of induction, reflecting the very rapid and complete induction observed in the majority of animals. Precise regulation of the MIC transgene was evident based on the concurrent observations that rtTA/MIC mice developed mammary tumours exclusively and that all control animals (both induced and un-induced) remained tumour-free after one year post-induction.

Tumour growth in rtTA/MIC mice progressed differently from what has been observed in the MMTV-PyV mT model. In the latter, tumours develop as focal masses in each gland that are easily measurable. At defined time-points, histological analysis of the inguinal mammary glands from MMTV-PyV mT mice shows a gradient of transformation, with the older and more advanced lesions proximal to the nipple, and newer lesions at earlier stages of tumourigenesis towards the terminal end buds of the epithelial network [13]. While distinct masses are initially palpable in an induced rtTA/MIC mouse, the entire gland promptly thickens within days, making it difficult to perform calliper measurements at this stage. Animals sacrificed at onset (approximately four days post-induction) or two weeks post-induction harboured inguinal mammary glands filled with early lesions (data not shown; Additional file 2: Figure S2). This difference between the two models could be explained by the fact that constitutive PyV mT-mediated transformation occurs during puberty as the ductal epithelial network progressively penetrates the fat pad, while in the MIC model transformation was initiated in an almost mature gland.

All tumour-bearing rtTA/MIC females were sacrificed at a total tumour volume of approximately six cubic centimetres (denoted as “end-stage”). Histological analysis of mammary glands and tumours from these animals revealed the presence of all previously characterized stages of PyV mT tumourigenesis, ranging from hyperplasia, to MIN/adenoma, and finally to early and late carcinoma (Figure 1B; Additional file 3: Figure S3B). Adjacent mammary gland whole mounts from tumour-bearing mice were also fully transformed (Additional file 3: Figure S3A). Mammary gland sections and whole mounts from age-matched control animals were normal (Figure 1B; Additional file 4: Figure S4). It appeared that our novel inducible PyV mT strain was closely recapitulating the histological stepwise tumour progression documented in the MMTV-PyV mT model.

rtTA/MIC mammary tumours co-express the PyV mT oncogene and a functional Cre recombinase

Having established that mammary tumours were indeed inducible in the rtTA/MIC system, our next step was to verify expression of the MIC transgene by immunohistochemistry. PyV mT and Cre recombinase antibodies stained the membrane and nuclei, respectively, of cells in rtTA/MIC lesions in a mosaic pattern (Figure 2A). Notably, normal ductal epithelium in both age-matched controls and wildtype animals did not stain positively for PyV mT or Cre recombinase.

To confirm MIC transgene expression by immunoblot, protein extracts were prepared from mammary glands and tumours from rtTA/MIC mice sacrificed at palpation or at end-stage tumour burden (“late onset” refers to palpation after 16 days of induction). These lysates were positive for PyV mT expression; Cre recombinase was also detected in rtTA/MIC tumours, although it was lowly expressed in mammary gland lysates which likely have relatively less epithelial content as indicated by E-cadherin levels (Figure 2B). Induced and un-induced control mammary glands did not express PyV mT or Cre recombinase protein.

In order to use this model for Cre recombinase/LOXP-mediated excision of genes, we needed to confirm that the Cre recombinase produced from the MIC transgene was functionally active. To accomplish this, we utilized a Rosa26-β-galactosidase reporter strain (“GTRosa”) in which the lacZ gene is downstream of a LOXP-flanked STOP cassette [12]. The presence of Cre recombinase allows for expression of the β-galactosidase enzyme from the lacZ transgene, which can cleave the compound X-gal into an insoluble blue-coloured product. The mammary epithelium of rtTA/MIC/GTRosa tumour sections turned blue upon staining with X-gal, indicating that Cre recombinase had been expressed and active in these cells (Figure 3). This outcome is comparable to an MMTV-PyV mT/MMTV-Cre recombinase/GTRosa mammary tumour in which there are no conditional alleles present. Collectively, these results demonstrate that the rtTA/MIC mouse model can be used to study Cre recombinase-dependent genetic alterations in conjunction with PyV mT oncogenic activation.

The rapid induction of rtTA/MIC lesions is associated with metastatic dissemination of tumour cells to the lung

One of the most useful features of the original constitutive PyV mT model is the ability of the mammary tumours to effectively metastasize to the lungs, an organ that is a common site of distal lesions in the human disease. To determine if rtTA/MIC mammary tumours were capable of forming pulmonary metastases, we examined sections of the lung lobes from animals that had reached similar end-stage tumour burdens. All tumour-bearing mice presented with lung metastases, albeit to varying degrees, with some lungs harbouring only a few small lesions while others were made up almost entirely of secondary tumour tissue (Figure 4A, B). These lung lesions stained positively for PyV mT, confirming that they derived from the primary rtTA/MIC mammary tumour (Figure 4C). Interestingly, there was no correlation between the extent of metastasis and tumour burden, extending the idea that PyV mT tumours are heterogeneous in their transforming capabilities and may thus also be in terms of malignancy (data not shown; [13]). Taken together, these observations demonstrate that this inducible MIC model reproduces many of the pathological features of the original MMTV-PyV mT strain.

De-induction of the MIC transgene results in immediate tumour regression and eventual recurrence of doxycycline-independent masses

Another important feature of inducible systems is the capacity to “turn off” the oncogene by withdrawal of the inducing agent; one can then evaluate whether tumours regress and if they have the potential to recur in the absence of transgene expression. To test this in our model, doxycycline treatment was discontinued for a cohort of rtTA/MIC mice bearing end-stage mammary tumours. Upon de-induction of PyV mT expression the tumours began to shrink rapidly (Figure 5A). By 10 weeks post-de-induction, most of the tumours had regressed to palpable masses that were no longer measurable. Interestingly, all of the de-induced rtTA/MIC mice eventually developed recurrent masses (15 to 50 weeks post-de-induction) and were sacrificed at burden endpoint. The number of measurable recurrent tumours arising was significantly less than the number of measurable masses the animal had prior to de-induction (Figure 5B). This suggests that the emergence of more focal, doxycycline-independent tumours in post-regression mice is a spontaneous event, in contrast to the consistent and complete penetrance of multifocal, doxycycline-dependent tumours driven by the inducible MIC transgene to all mammary glands of a given animal. Doxycycline-independent tumours arose only in rtTA/MIC mice and not in control animals de-induced at the same time, which remained normal (data not shown). Analysis of sections from mid-regression tumours and completely regressed mammary glands revealed relatively normal ductal structures surrounded by extensive stromal deposition and occasionally abnormal adipose tissue, which may explain why these glands remained palpable so long after de-induction (data not shown).

The morphology of the recurrent masses exhibited striking intra- and intertumoural heterogeneity, with individual tumours differing among animals and among tumours from the same mouse (Figure 6A; Additional file 5: Figure S5A). Some histopathologies resembled pre-regression adenocarcinoma while others were more divergent, such as epithelial-mesenchymal-transition (EMT)-like and striated/punctate morphologies. All mice with recurrent tumours had evidence of lung metastases at sacrifice, varying in number, size and stage (data not shown). A single animal sacrificed prior to recurrent tumour development did not present with any metastases which suggests that, at least in this individual case, MIC-induced lung lesions may regress in parallel with the primary tumour upon doxycycline withdrawal.

To determine if the recurrent tumours were no longer dependent on the PyV mT oncogene, we subjected lysates from the five tumours shown in Figure 6A to immunoblot analysis (Figure 6B). Two samples clearly showed detectable bands at the expected size for PyV mT (2376 R1 and 3027 L4), while weaker signals were observed in two other samples (2380 L1 and 2692); one sample appeared to be completely negative for PyV mT expression (2379). To achieve a more quantitative assessment of PyV mT levels, qRT-PCR was carried out on these and other recurrent tumours as well as corresponding metastatic lung lesions from two animals. We found that PyV mT transcript levels in the recurrent tumours reflected the protein levels obtained by immunoblotting (Additional file 5: Figure S5B). For the most part, re-expression of the PyV mT transgene correlated with the incidence of adenocarcinoma in the tumour’s corresponding histological section. The lack of complete correlation can be explained by the fact that we must use different pieces of a histologically heterogeneous tumour for different analyses. Interestingly, a PyV mT transcript was detected in a metastatic lung lesion in addition to a recurrent tumour from the same animal (2376; Additional file 5: Figure S5B); PyV mT protein expression in the lung lesions was confirmed by immunohistochemistry (data not shown). On the other hand, in an animal with a PyV mT-expressing recurrent tumour (2380 L1) and a -non-expressing recurrent tumour (2380 L3), PyV mT transcript was undetectable in a metastatic lung lesion. This may reflect the presence of separate recurrent lesions that either re-expressed the transgene or arose in the absence of PyV mT transcript. It may additionally indicate the capability of non-PyV mT-expressing, recurrent mammary lesions to metastasize to the lungs, or the PyV mT-independent recurrence of an originally doxycycline-dependent lung metastasis in situ. We should note that the lower transcript levels of PyV mT transgene in the rtTA/MIC tumours relative to that of the MMTV-PyV mT tumour are likely due to the different strengths of the promoters driving the transcription of the oncogene (TetON versus MMTV). This does not impact on overexpression of PyV mT protein in the inducible system, as evidenced by the strong levels detected by immunoblot (Figure 2B) and ultimately the fact that transformation occurs rapidly in the model.

It appeared that while some recurrent tumours arose by re-expressing the PyV mT transgene, others did so by alternative mechanisms. In an attempt to identify these potential mediators of recurrence, we used an antibody array to analyse phospho-RTK levels in primary and recurrent tumour lysates and focused on candidates with relatively high fluorescence intensities, specifically epidermal growth factor receptor (EGFR), ErbB2 and platelet-derived growth factor receptor (PDGFR) β (Additional file 6: Figure S6A). While not all candidates validated by immunoblotting, we did observe phosphorylation of both ErbB2 and PDGFRβ in at least some of the recurrent tumours, suggesting that RTK signalling was actively occurring (Additional file 6: Figure S6B). ErbB2 is known to be up-regulated during mammary tumour progression in the MMTV-PyV mT model [4]. While we observed relatively low expression of ErbB2 in our samples as compared to an MMTV-NIC control lysate, the strong levels of phosphorylated ErbB2 in the doxycycline-independent rtTA/MIC recurrences may represent an avenue of recapitulating the signalling associated with PyV mT tumourigenesis in the absence of transgene re-expression. This may be the case for 2379, which also showed overexpression of c-Myc protein; interestingly, amplification of the c-Myc gene has been observed in a model of recurrence after de-induction of the doxycycline-dependent oncogene [14].

To further analyze the potential occurrence of cooperating oncogenic events during the process of doxycycline-independent recurrence of rtTA/MIC mammary tumours, we sequenced regions of the three Ras genes (Hras, Kras1 and Nras) and of Trp53 that are orthologous to those frequently mutated in human cancers. Notably, mutations in these genes have been previously identified as potential driving events in the recurrence of other doxycycline-driven transgenic mouse tumour models [15, 16]. No mutations were found in any of the genes examined in doxycycline-dependent rtTA/MIC mammary tumours (data not shown). In recurrent mammary tumours, we found no mutations in exons 2 and 3 (containing codons 12 and 61) of either of the Ras genes but did identify an arginine-to-cysteine mutation at residue 245 of Trp53 (R245C) in one recurrent mammary tumour (2380 L1; data not shown). The affected residue corresponds to R248 of human TP53, which is frequently mutated in human cancer. This result suggests that mutations in known tumour suppressor genes can occur in recurrent rtTA/MIC mammary tumours. However, at least in the case of Trp53, they may be relatively infrequent (1/10 samples examined). A more comprehensive mutational analysis (for example, using exome sequencing) of doxycycline-dependent and recurrent rtTA/MIC mammary tumours could be undertaken in the future to provide additional information on cooperating genetic events during tumour recurrence.

Collectively, these data illustrate that, while we can demonstrate rapid tumour regression in rtTA/MIC animals by withdrawal of doxycycline, the emergence of doxycycline-independent tumours ultimately transpires. This can be attributed in at least some cases to the reactivation of the PyV mT transgene and corresponds with an adenocarcinoma phenotype. In other cases, tumour recurrence may be associated with activation of RTK signalling and/or cooperating oncogenic mutations, such as the observed mutation in Trp53. These events may correlate with a different spectrum of tumour histopathologies, since the occurrence of the R245C mutation in 2380 L1 correlates with the appearance of an EMT-like morphology in addition to adenocarcinoma (Figure 6A). This is in keeping with the established tendency of Trp53 mutations to induce tumours with EMT-type histopathological features in transgenic mouse models [17].