Modeling the MYC-driven normal-to-tumour switch in breast cancer

The potent MYC oncoprotein is deregulated in many human cancers, including breast carcinoma, and is associated with aggressive disease. To understand the mechanisms and vulnerabilities of MYC-driven breast cancer, we have generated an in vivo model that mimics human disease in response to MYC deregulation. MCF10A cells ectopically expressing a common breast cancer mutation in the PI3 kinase pathway (PIK3CAH1047R) lead to the development of organized acinar structures in mice. However, expressing both PIK3CAH1047R and deregulated-MYC lead to the development of invasive ductal carcinoma, thus creating a model in which a MYC-dependent normal-to-tumour switch occurs in vivo. These MYC-driven tumors exhibit classic hallmarks of human breast cancer at both the pathological and molecular levels. Moreover, tumour growth is dependent upon sustained deregulated MYC expression, further demonstrating addiction to this potent oncogene and regulator of gene transcription. We therefore provide a MYC-dependent human model of breast cancer which can be assayed for in vivo tumour initiation, proliferation, and transformation from normal breast acini into invasive breast carcinoma. Taken together, we anticipate that this novel MYC-driven transformation model will be a useful research tool to both better understand MYC’s oncogenic function and identify therapeutic vulnerabilities. Conflict of interest statement The authors declare no potential conflicts of interest.


INTRODUCTION
MYC deregulation occurs in the majority of human cancers and is therefore an ideal therapeutic target. 1 Deregulation can be broadly defined as any genomic, epigenetic or signaling pathway aberration that results in sustained and often elevated MYC expression and activity.
However, targeting this potent oncoprotein has been difficult using traditional approaches as MYC is a transcription factor lacking enzymatic pockets. Alternative strategies are therefore under intense investigation, including targeting MYC:DNA interaction with small protein inhibitors such as Omomyc or ME47. 2,3 In addition, identifying the unique vulnerabilities which arise in a MYC-driven tumour can reveal targetable synthetic lethal interactions. [4][5][6][7][8] There is therefore a need for a diverse set of MYC-driven models to identify and test new therapeutic strategies in a MYC-transformed setting.
Many approaches have been developed to model MYC in cancer, including cancer cell lines and genetically engineered mouse models (GEMMs). Rodent models include the widely used Rat-1A cell line, the Eμ-MYC transgenic model of B-cell lymphoma and the MMTV-MYC model of breast cancer (BC). [9][10][11][12] Additionally, human MYC-dependent models, such as the Burkitt lymphoma cell line model P493-6 with regulatable MYC levels, have been used extensively to interrogate transcriptional regulation by MYC. [13][14][15] Moreover, the MycER chimeric protein system, which can be applied to cell line and mouse models of many cancer types, has been used for studies of MYC in the cell cycle, genomic instability, transformation and synthetic lethal interactions. [16][17][18][19][20][21][22] These models have provided several new insights into MYC biology, suggesting further development of diverse models will also provide important contributions, particularly if they are complementary and obviate associated limitations with existing models. For instance, GEMMs are time-consuming to manipulate genetically, are resource-intensive and often do not accurately model all human disease at the histological level.
Specifically, when modeling human breast cancer, there are numerous differences between mouse and human mammary gland development and the surrounding stroma that need to be considered. 23 In addition, most cancer cell line models do not accurately reflect human tumours in vivo, are derived from already fully transformed tumour tissue, and are not necessarily driven by or dependent on deregulated MYC. We have therefore set out to design a functionally complementary MYC-driven transformation model that meets the following criteria: is fullytransformed in response to, and dependent-upon, deregulated MYC; recapitulates human disease at both the pathological and molecular levels in vivo; and, is easy to manipulate genetically.
These criteria outline a model that best reflects the complexity of human tumours in vivo while ensuring that identifying MYC-dependencies is possible. To this end, we used a non-transformed parental cell (MCF10A) and developed an isogenic panel of cell lines that could meet these standards and model the transformation of BC upon the introduction of deregulated MYC. We evaluate MYC-dependency in vivo and characterize the histology and mRNA expression changes that occur in the these MYC-driven tumours.

RESULTS and DISCUSSION
In vivo transformation of MCF10A cells is dependent on active PI3K signaling, deregulated MYC, and requires MBII.
We chose to model BC as MYC is often deregulated and contributes to the progression of this disease. 24,25 Recent large whole-genome sequencing studies of BC patient tumours revealed that the most frequently mutated and amplified oncogenes in BC are PIK3CA and MYC, respectively. 26 PIK3CA is mutated in approximately 30% of BC patient tumours, whereas MYC deregulation, primarily scored through amplification, occurs in approximately 20% of BCs. 26,27 To model BC transformation, we sequentially introduced these commonly mutated oncogenes into the MCF10As, an immortal, non-transformed breast cell line. 28 MCF10As were chosen due to their inherently stable genome, which ensures the oncogenic conversion into the transformed state is due to the introduction of ectopic oncogenes rather than an indirect consequence of genomic instability and the selection of a transformed clone. In addition, the MCF10As are a relevant model as the few mutations acquired during spontaneous immortalization are prevalent in human BC, including CDKN2A deletion. 29 Finally, despite harbouring a focal amplification at chromosome 8q24, total MYC levels remain highly regulated, decreasing rapidly in response to anti-proliferative conditions which is a hallmark of normal MYC regulation in non-transformed cells. 29,30 Thus, the MCF10A cells fulfilled our criteria of a parental breast cell system to base our model.
We first stably introduced the empty vector control or the activated PIK3CA H1047R allele into MCF10A cells. The latter was active in these cells as AKT phosphorylation at serine 473 was readily detectable ( Figure 1A To assay for transformation, we cultured our isogenic cell line panel as breast organoids on Matrigel to model three-dimensional acini formation in vitro ( Figure 1C). 31 10A.EE and 10A.EM cells formed round, organized acinar structures consistent with previous publications. 30 In contrast, 10A.PE and 10A.PM cells formed large and disorganized acinar structures, demonstrating a transformed phenotype driven by PIK3CA activation in vitro, but not by MYC deregulation alone (Supplementary Figure S1D). Importantly, breast organoids are grown in specialized culturing conditions and cells undergo polarization programs while interacting with matrigel, whereas standard tissue culture lack these variables. These differences are likely important for the dramatic phenotypic differences observed.
Considering the importance that environmental context has on the phenotypes observed, we progressed to evaluate xenograft growth in vivo. This xenograft assay recapitulates the tumour microenvironment and offers an increasingly relevant context to study tumour formation and growth. To assay xenograft growth, we generated cells expressing PIK3CA H1047R and then  Figure S1E). The latter serves as a negative control as MBII is evolutionarily conserved and required for MYC-dependent transformation. 9,32 10A.PE cells formed small but measurable masses in vivo, unlike parental MCF10As and 10A.EM cells, which did not establish any mass (Supplementary Figure S1F). This suggests that activation of the PI3K signaling pathway is required for the establishment of MCF10A xenografts, reflective of the in vitro organoid data suggesting that these cells are more transformed. By contrast, 10A.PM cells were able to form significantly larger masses compared to 10A.PE cells ( Figure 1D), which was surprising as MYC did not potentiate transformation of 10A.PE cells in our in vitro organoid assays ( Figure 1C). Cells expressing ΔMBII (10A.PΔMBII) formed small palpable masses, similar to 10A.PE, demonstrating that tumour xenograft growth was dependent on MYC and the conserved MBII region ( Figure 1D). Doubling times were 6 days for 10A.PM and 60 days for 10A.PE and 10A.PΔMBII xenografts. Histological examination showed that 10A.PE and 10A.PΔMBII xenograft masses were composed of acinar structures embedded within an extracellular matrix, strikingly similar to the histology of normal human breast acini ( Figure 1E).
By contrast, 10A.PM xenografts possessed major phenotypic differences, appearing as dense, fully-transformed breast tumours with features reminiscent and highly similar to human disease.
While active PI3K signaling transforms breast organoids in vitro, both PIK3CA and deregulated MYC are required for transformation in vivo.

Deregulated MYC transforms breast acini into invasive ductal carcinoma in vivo.
In clinical practice, tumours are evaluated through, histology, immunohistochemistry (IHC) and/or molecular profiling to characterize tumour subtype and features of aggressive disease. We applied these analyses to our xenografts, in a blinded-manner, with the assistance of a practicing BC pathologist. 10A.PM and 10A.PΔMBII growths had significantly increased Ki67 staining compared to 10A.PE whereas TUNEL staining was low in all samples ( Figure 2A).
Interestingly, while 10A.PΔMBII tumours had an increase in Ki67 signal, the overall tumour volume was equal to 10A.PE xenograft nodules. This suggests that MYC ΔMBII is able to confer some proliferative advantage due to retention of partial MYC-related activity, but is not able to drive transformation, which is consistent with previous reports. 9, 32 We also analyzed clinically relevant markers such as the estrogen and progesterone receptors (ER and PR), human epidermal growth factor receptor 2 (HER2), and cytokeratin 5 (CK5) as a marker of basal cells ( Figure 2B, left). The MCF10A cell line has been previously characterized as basal and triple-negative for 8 breast cell receptors in vitro. 33 Introduction of PIK3CA H1047R and MYC maintained the basal and triple-negative state in vivo.
We then used clinical markers to distinguish the degree of transformation within each xenograft. Malignant disease can be categorized as either carcinoma in situ (CIS) or invasive ductal carcinoma (IDC) and is characterized by transformed cells contained within or invading through a layer of myoepithelial cells surrounding the acinar structure, respectively. To assess these conditions, we stained for myoepithelial cells using two markers, P63 and SMA. 10A.PE and 10A.PΔMBII acini were surrounded by myoepithelial cells and were importantly identified with human-specific epidermal growth factor receptor (EGFR) antibody, indicating that these cells were of MCF10A origin. However, 10A.PM tumours lacked p63 and SMA co-expression, suggesting a lack of myoepithelial cells in MYC-driven tumours. Our pathology report determined that a majority of 10A.PE growths were normal breast acinar structures with few cases of atypical hyperplasia (AH) ( Figure 2B, right). 10A.PΔMBII growths developed moderately more AH lesions than 10A.PE xenografts, despite MBII being essential for tumour growth. These data align with our Ki67 proliferation data and previously published results showing MBII is required for transformation but can retain some MYC-related activity in tissue culture assays. 9,32 By contrast, 10A.PM tumours, which were highly proliferative and lacked myoepithelial cells, were scored as high-grade IDC. Therefore, the addition of deregulated MYC in this in vivo model system transforms cells into IDC in an MBII-dependent manner. Thus, we are able to evaluate MYC-driven tumour progression (10A.PM) against a non-transformed control (10A.PE) in vivo, as well as measure tumour volume, and evaluate tumour pathology; a rich dataset that to the best of our knowledge uniquely distinguishes this model from previous solid human MYC-driven cancer models.

10A.PM tumours have MYC and invasive ductal carcinoma expression signatures.
Our histological analysis demonstrates that the in vivo 10A.PE and 10A.PM isogenic tumour pair represents normal breast acini and IDC, respectively. We took this opportunity to identify an in vivo MYC-dependent transformation expression dataset. We re-established 10A.PE and 10A.PM xenografts and isolated RNA from 3 biological replicates for RNA sequencing (RNA-seq) ( Figure 3A). Contaminating mouse RNA was removed using Xenome, 34 leaving 5000 significantly differentially expressed genes (Supplementary Table S1). We performed gene set enrichment analysis (GSEA) 35 Figure 4B). This data demonstrates that this model is both driven by, and addicted to, deregulated MYC in a reproducible and quantitative manner. Furthermore, we show that targeting the deregulated fraction of MYC is sufficient to inhibit tumour growth as endogenous MYC expression is not affected by DOX treatment ( Figure 4A). Therefore, the observed phenotypes, signaling pathways, and vulnerabilities that arise in vivo are a consequence of the regulatable and ectopic allele of MYC and importantly, are not due to permanent genomic or microenvironmental changes after deregulated MYC-driven transformation.
The isogenic MCF10A panel presented here offers unique advantages and is complementary to current MYC-dependent models (Supplementary Figure 1F). [37][38][39] Interestingly, the observed MYC-driven phenotype was exclusive to the in vivo xenograft assay.
This highlights the importance of identifying a relevant context when studying MYC-dependent transformation, as our in vitro studies demonstrated little oncogenic MYC activity towards proliferation or in vitro acini transformation, and in no way was able to predict the striking in vivo response. Moreover, studying cancer biology using in vivo models most accurately reflects human disease and incorporates the contribution of tumour microenvironment and cellular stress.
The in vivo context is therefore ideal for studying deregulated MYC and identifying MYCdependent vulnerabilities as the relevant stress-signaling pathways will be active.
We have demonstrated that this model can recapitulate human BC at both the molecular and pathological level, and is transformed by deregulated MYC. We anticipate that this model will be a useful tool for identifying the context-dependent functions and vulnerabilities of MYCdriven cancers, that are more difficult to discover using traditional in vitro or murine models. Additionally, this system is reproducible and scalable; ideal characteristics for the identification and validation of anti-MYC therapies. Indeed, we have recently demonstrated the utility of this model through the validation of the G9a histone methyltransferase as an epigenetic, therapeutic target in MYC-driven cancers. 40 We anticipate that this system will be able to predict the efficacy of novel therapeutic drugs targeting MYC and will identify novel vulnerabilities of MYC-dependent BCs.

METHODS
Cell culture. MCF10A epithelial cells were a kind gift of Dr. Senthil Muthuswamy and were cultured in MCF10A culture media, as previously described. 31 Cell culture images were taken with a 32X(Ph) objective on an AxioObserver microscope. EdU incorporation assay. The MCF10A isogenic panel was grown for 24 hours in fully supplemented media and allowed to reach 50% confluency. At 50% confluency, cells were treated with EdU for 4 hours, processed according to manufacturer's instructions (Click-iT EdU Alexa Fluor 488 Flow Cytometry Assay Kit, Cat# C10420, Life Technologies) and analyzed for EdU positivity using the LSR II flow cytometer (BD Biosciences) and FlowJo v10 software.
Three-dimensional Matrigel morphogenesis assay. Cells were cultured on Matrigel (Cat# 354230, Corning) and in specialized MCF10A media for three-dimensional cultures, for 12 days as previously described. 31 Images were acquired on day 12 using a FLUAR 5x/0.25 NA lens on an AxioObserver microscope (Zeiss). Cell line groups were blinded and quantified based on qualitative assessment of acinar circularity and size. Small round acini were considered normal; large and disorganized acini were considered transformed. One-way ANOVA with Bonferroni post-test for multiple testing was used for analysis.
Xenografts. 1x10 6 cells were suspended in 50% matrigel (Cat# 354262, Corning) to a final volume of 0.2 mL and injected subcutaneously into the flanks of female NOD-SCID mice. For Tetracycline repressed cell lines, 100 μg/mL DOX (Sigma) was supplemented into the drinking water, prepared fresh twice a week. One-way ANOVA with Bonferroni post-test for multiple testing was used for analysis.
Immunohistochemistry. Extracted tumours were fixed in 10% buffered formalin (Sigma) for 24-48 hours at room temperature and then stored in 70% EtOH at 4°C. Samples were then paraffin embedded, sectioned at 4 μm thickness and stained with hematoxylin and eosin. Ki67 following the manufacturer's protocol. RNA was sent to the Princess Margaret Genomics Centre and sequenced using the Illumina NextSeq 500 for an average of 40M reads per xenograft growth. Fastq files were filtered using Xenome in order to filter out mouse contamination. Only reads mapping exclusively to the mouse reference were filtered out. The remaining reads were mapped to the human reference genome (hg19) and read counts estimated using STAR.
14 Expressed genes were defined as requiring at least 10 reads in at least 2 samples. Read counts were upper quantile normalized. Finally, EBSeq was used to test for differential expression between experimental conditions. Genes were determined to be differentially expressed given an  (A) Tumours harvested in Figure 1C were stained (left) for Ki67 and TUNEL and were quantfied (right) for the degree of Ki67 and TUNEL staining. Individual qunatifications per tumour are shown, *P<0.05, ***P ≤ 0.001, one-way ANOVA with Bonferroni post-test for multiple testing. Scale = 100μm. (B) Tumours harvested in Figure 1C were stained (left) for ER, PR, HER2, CK5, EGFR, p63 and SMA by the PRP laboratory. A pathology report (right) was produced for each tumour using the combination of H&E as shown in Figure 1D and IHC markers used in this study. Scale = 100μm.      Figure 1C were stained (left) for Ki67 and TUNEL and were quantfied (right) for the degree of Ki67 and TUNEL staining. Individual qunatifications per tumor are shown, *P<0.05, ***P ≤ 0.001, one-way ANOVA with Bonferroni post-test for multiple testing. Scale = 100μm. (B) Tumors harvested in Figure 1C were stained (left) for ER, PR, HER2, CK5, EGFR, p63 and SMA by the PRP laboratory. A pathology report (right) was produced for each tumor using the combination of H&E as shown in Figure 1D and IHC markers used in this study. Scale = 100μm.