There are now drugs to treat many types of cancer, but questions still remain around how these diseases start in the first place. Researchers think that tumor growth begins when a single cell suffers damage to certain sites in its DNA that eventually cause it to divide uncontrollably. That damaged cell, and its descendants, go on to form a lump, or tumor. The trouble with proving this theory is that it is hard to watch it happening in real time. Doctors usually only meet people with cancer when their tumors start to cause health problems. By this point, the tumors contain millions of cells.

A way to watch the very beginnings of a cancer could reveal risk factors within a tissue that foster the growth of a tumor. But first, researchers need to test their theory about how the disease begins in the first place. One way to do this is to surround a single cancer cell with healthy cells and watch what happens next. To do this, Alladin, Chaible et al. took healthy cells from the breast tissue of mice and grew them in the laboratory into mini-organs called organoids. These organoids share a lot of features with actual mouse breast tissue; they can even make milk if given the right hormones.

Once the organoids were ready, Alladin, Chaible et al then started modifying a small number of single cells inside them by switching on genes called oncogenes, which are known to drive cancer formation in humans. Using fluorescent proteins and a sheet of laser light it was possible to watch what happened to the cells over time. This revealed that, even though all the oncogene-driven single cells received the same signals, not all of them started to divide uncontrollably. In fact, a single modified cell had a low chance of forming a tumor on its own. The more oncogene-driven cells there were near to each other, the more likely they were to form tumors. Alladin, Chaible et al. think that this is because the healthy tissue interacts with the modified, oncogene-driven cells to suppress tumor formation. It is only when a larger number of modified cells group together and start to communicate with each other that they can override the inhibitory messages of the healthy tissue.

How healthy tissue stops single modified cells from forming tumors is not yet clear. But, with this new mini-organ system, researchers now have the tools to investigate. In the future, this could lead to new strategies to stop cancer before it has a chance to get started.

Organoid cultures grown from cell-lines or primary cells have been successfully employed to study molecular mechanisms during different stages of tumorigenesis (Clevers, 2016; Havas et al., 2017; Simian and Bissell, 2017). However, they are derived from primary material that usually allows oncogenic activation in all cells of the tissue and therefore cannot reproduce the localized expansion at a defined part of the tissue that is seen in the patient situation. To study tumor initiation, models need to be established wherein only a few tumorigenic cells expand in the context of its immediate non-tumor microenvironment (Tabassum and Polyak, 2015) and the interplay of these populations can be visualized.

Long-term imaging of primary organoids has been achieved via light-sheet microscopy (Drost et al., 2015; Serra et al., 2019; Verissimo et al., 2016) and these efforts have benefited from the lower phototoxicity provided by SPIM imaging. However, regarding fast dividing cells in tumor organoids, tracking single-cell dynamics necessitates high resolution imaging which in turn limits the time frame in which organoids can be imaged without phototoxic effects (Held et al., 2018). Conversely, imaging primary organoids for longer time periods requires an offset of temporal and cellular resolution that eventually cannot allow single-cell fate tracking (Dekkers et al., 2016).

Two models have been used to understand tumor progression and heterogeneity (Plaks et al., 2015); the (i) hierarchical model refers to tumor propagating cells as cancer stem cells (CSCs) (Kreso and Dick, 2014) and the (ii) stochastic model states that every cell within a tumor is equally likely to be the cell of origin and facilitate tumor initiation and progression (Vogelstein et al., 1988; Williams et al., 2007). These models for tumor propagation can be reconciled when considering that the population of oncogene driven cells have the capacity to interconvert between differentiated- and stem-like states (Chaffer and Weinberg, 2015; Gupta et al., 2011; Plaks et al., 2015), a flexibility that is supported by the microenvironment (McGovern et al., 2009; Quail et al., 2012). These concepts, however, have not been probed in the context of tumor initiation upon stochastic oncogene activation in an established cell-layer.

Here we present a novel model of breast tumorigenesis where only single cells express oncogenes within healthy mammary acini and are able to establish localized outgrowth. We thereby overcome the above-mentioned limitations of studying tumorigenesis events in the face of tissue-wide transformation. Furthermore, we perform long-term imaging of these stochastic breast tumor acini for the first time at a temporal resolution that allows us to follow single-cell fates. We also integrate this approach with an image analysis pipeline capable of segmenting cells in their dynamic progression towards tumorigenesis, so they can be tracked individually over time.

For modeling tumorigenesis in breast tissue, we use an inducible mouse model of breast cancer (TetO-MYC/TetO-Neu/MMTV-rtTA transgenic mice) (Fry et al., 2016; Moody et al., 2002; Podsypanina et al., 2008) that has been shown to recapitulate hallmarks of human breast disease (Havas et al., 2017; Jechlinger et al., 2009; Figure 1a). In this tractable transgenic mouse model, the activity of two potent oncogenes – MYC and Neu (the rodent homolog for the human ERBB2 gene used in this mouse model)– are under the control of a Tet-O promoter that can only be activated in the presence of both the rtTA (reverse tetracycline-controlled transactivator) protein and the doxycycline compound. This three-part inducible system allows for temporal control of oncogenic expression by regulating doxycycline in the medium or animal diet and spatial control by the MMTV promoter that confines expression of the rtTA protein to the cells of the mammary lineage (Bockamp et al., 2002). We modify this tissue wide tumorigenesis model (tri-transgenic (T) model; Figure 1a, upper panels) to generate a stochastic system by retaining only the oncogenic constructs, making them TetO-MYC/TetO-Neu mice (bi-transgenic (B) model; Figure 1a, lower panels). The rtTA inducer gene is then lentivirally delivered to single cells within luminal epithelial acini derived from the bi-transgenic (B) mouse glands, granting only this subset of cells the competence to express transforming oncogenes and thereby preventing tissue wide transformation. Hence, this novel (B) model should allow acini to present with two cell populations –tumorigenic and non-tumorigenic– and could be used to study the dynamics of early tumorigenesis in an otherwise normal epithelium (Figure 1a, lower right panels).

Figure 1 with 1 supplement see all

Download asset Open asset

Primary mammary epithelial cells derived from transgenic mice were seeded in 3D matrigel as single cells to be grown for 48 hr to form small acini consisting of around 15 cells that arrange around a small lumen. A small number of single cells in these acini were then transduced with lentiviral particles (Figure 1a, middle panel). In the tissue-wide tumorigenesis model(T), acini were transduced with the reporter virus (pLv-pGK-H2B-GFP) that marks a subset of cells with H2B-GFP, while tissue wide rtTA expression is driven in all cells (Figure 1a, left panels). To achieve stochastic tumorigenesis, bi-transgenic (B) acini were transduced with the inducer-reporter virus (pLv-pGK-rtTA-p2A-H2B-GFP) that expresses rtTA and reporter H2B-GFP in only these single cells within the normal epithelium (Figure 1a, right panels). Then, doxycycline was supplemented in the medium to induce tumorigenic growth in rtTA expressing cells. Immunofluorescent staining of 3D matrigel cultures, for both sets of doxycycline-induced transduced acini, was used to validate transgene specific protein expression of the MYC oncogene (human) in only the transduced cells of (B) acini as opposed to all cells of (T) acini (Figure 1b, anti-human MYC specific antibody). We FACS sorted H2B-GFP marked cells from both systems respectively for qPCR analysis of transgene specific MYC and Neu mRNA expression. This way we normalized doxycycline dosage in both systems to obtain the same mRNA expression levels of oncogenes using 800 ng/ml doxycycline in the tissue wide (T) system and 600 ng/ml doxycycline in the stochastic (B) system (Figure 1—figure supplement 1).

To verify oncogene induction patterns and to understand the impact on epithelial morphology in the acini, 3D matrigel cultures of both doxycycline-induced (T) and (B) transduced acini were analyzed by immunofluorescent staining for polarity marker distribution. Lentivirus-transduced structures from the tissue wide induction system (T) presented with filled lumens that contained both infected (GFP positive) as well as non-infected oncogene-expressing cells. These structures exhibited overall disrupted epithelial polarity and some random microlumen, as reported previously (Jechlinger et al., 2009; Figure 2a). For the stochastic tumor induction system (B), we observed 2 phenotypes after 96 hr of oncogene expression; some of the (B) acini showed exclusively H2B-GFP marked cells in hyperplastic areas that also displayed double cell rim morphology and disturbed epithelial polarity (Figures 2a and 96 hours ON DOX, left panels), while the other (B) acini did not show expansion of the few H2B-GFP positive cells (Figures 2a and 96 hours ON DOX, right panels). Intrigued by these 2 distinct phenotypes, we analyzed transduced (H2B-GFP expression) (B) acini for their oncogenic MYC expression (Figure 2b). We verified that multi layered regions in acini contained mainly H2B-GFP positive cells that expectedly stained for MYC (Figure 2b, middle panels). For normal appearing regions, this analysis showed that the inability of single H2B-GFP positive, transduced cells to expand over 96 hr was not due to the lack of MYC expression (Figure 2b, lower panels).

Figure 2

Download asset Open asset

To follow up these observations in more detail over time, we bred the nuclear reporter H2B-mCherry into the T and B mice to mark all the cells in the acinus for inverted light-sheet microscopy (Luxendo InVi SPIM, Figure 3—figure supplement 1). The InVI SPIM was adjusted for non-phototoxic, long-term imaging (up to 4 days, every 10 min with 1 µm z-spacing). The T/H2B-mCherry acini transduced with reporter virus proliferated swiftly upon doxycycline addition showing expansion of both the marked and unmarked cells (Figure 3a); a sturdy tumor phenotype developed, manifested by multi-cell-layered rims and pronounced proliferation-associated-apoptosis in all acini (Figure 3—Video 1). In contrast, B/H2B-mCherry acini transduced with inducer-reporter virus, displayed phenotypic variation upon induction of oncogenes in the transduced cells. Some acini showed fast clonal expansion of oncogene-expressing cells that form multilayer clusters in the acinus rim. This proliferative phenotype seems to stem from several transduced cells in vicinity to each other at the start of time-lapse imaging (Figure 3b, upper panel; Figure 3—Video 2). Again, more sparsely infected acini, did not sustain proliferation of the oncogene-expressing cells (Figure 3b, lower panel; Figure 3—Video 3). Taken together, the phenotypes for (B) and (T) acini are consistent with the respective behavior observed with the 96 hr end point immunofluorescent analysis (Figure 2a and b), verifying these prior observations and excluding phototoxic effects due to longitudinal SPIM imaging.

Figure 3 with 6 supplements see all

Download asset Open asset

To analyze the dual-color light-sheet movies (H2B-mCherry—all cells in the acinus , H2B-GFP—transduced cells within the acinus) on a single-cell level, we developed a big image data compatible image analysis pipeline that allows efficient visualization of longitudinal big image data, nuclear segmentation and single-cell tracking in 3D (Figure 3c, Figure 3—figure supplement 3 and Figure 4—figure supplement 2). The Big Data Tools plugin (Big Data Processor (Tischer, Norlin, and Pepperkok)) for Fiji (Schindelin et al., 2012) was used to visualize the 2D image stacks recorded over time. It uses lazy loading to stream the image files, which can be up to a few terabytes in size per acinus over 3–4 days of imaging. It also allows for easy cropping, binning, chromatic shift correction between channels and file format conversion to Imaris compatible formats. Once the raw data have been pre-processed, the CATS plug-in from Fiji (Tischer and Pepperkok) was used for segmenting the cell nuclei in the green channel images. The CATS plug-in uses machine learning algorithms to predict the probabilities of all pixels in the image and classify them into pre-defined classes. Three classes — background, nucleus, boundary — were trained manually on the CATS tool and the segmented images were exported to the commercially available Imaris software (Bitplane AG, 2020, Software available at http://bitplane.com) for 3D rendering and tracking. The nucleus class channel from the CATS tool was used for surface rendering and the surfaces in all time points were tracked over the course of the movie. Cell tracking allowed us to follow the clonal evolution for each transduced cell in the acinus over 3 days. Single transduced cells within one acinus show a difference in proliferation and cell fate as indicated in representative tracks (Figure 4a).

Figure 4 with 2 supplements see all

Download asset Open asset

To better understand the parameters that positively affect a transduced cell in the stochastic tumorigenesis model to start proliferating and establishing a tumor within a normal epithelium, we extracted the center of mass coordinates of all cells in the acinus at the start of the imaging and represented them as dots in a 3D space: depicted as magenta for normal cells and green for transduced cells in Figure 4b (upper right panel). Equipped with the tumor formation success of each stochastically transduced acinus, we hypothesized that tumors originate from groups of independently transduced oncogene-expressing cells in closer vicinity within the acinus. To explore this in detail, we defined clusters of cells in the stochastic model as a group of oncogene-expressing cells that are closer to each other than to other oncogene-expressing cells of the same acinus. Notably, these clusters contain both oncogene-expressing cells as well as normal cells of the acinus. The cluster volume was described as the sphere located at the center of mass of a cluster with a diameter equal to the distance between the two farthest oncogene-expressing cells of the cluster. Next, we extracted 9 features that represent reasonable factors that could influence tumor establishment in the epithelium such as size of the acinus (represented by number of cells in the acinus), density of cells in the acinus, infection efficiency (represented by number of infected cells in the acinus), number of cells in a defined cluster volume, number of independently transduced cells in a cluster volume, distances between all cells in a cluster volume, distances between transduced cells in a cluster volume, fraction of transduced cells in the cluster volume and number of cell-cell contacts between transduced cells in a cluster volume (Figure 4b, left panel).

We analyzed 20 acini and all the transduced cells in these acini (n = 150) at the start of the imaging to ascertain the effect of the immediate non-transformed cell microenvironment on tumor cell proliferation. Since clusters of oncogene-expressing cells could be tracked over time, each cluster was associated with either a tumor outcome, or not. To identify which features were linked to this outcome, we chose a model selection approach, based on information theory. We fitted a logistic regression model for all possible linear combinations of features and selected the best model based on the Akaike information criterion (with correction for small sample sizes) (Calcagno and Mazancourt, 2010). In this model only one feature, the ‘number of transduced cells in the cluster’, can predict whether a cluster can form a tumor. Each additional transduced cell in a cluster increases the odds of this cluster forming a tumor by 9 (Figure 4b, lower right panel). Since other models within a few information criterion units of the best one can be considered equally good at predicting tumor formation, we also looked at features included in models within 3 information criterion units from the best one. Relative importance of features can be estimated by the sum of the information criterion values of the models in which each feature appears (Buckland et al., 1997). Computing relative importance across the selected models also revealed ‘number of transduced cells in the cluster’ as the most important feature, followed by ‘number of cells in a cluster’ and ‘number of contacts between transduced cells in a cluster’. However, the only feature with significant contribution to tumor formation is ‘number of transduced cells in the cluster’. It is important to note that this result does not rule out contribution from other factors which more targeted experiments may be able to reveal.

We also analyzed the effect of whole acinar features at the beginning and end of the imaging period to try and detect acinus-specific advantages, like acinus size, cell density and cell proliferative rate. Acini with transduced cell clusters that displayed a tumor outcome were compared to acini where the cell clusters did not result in tumor outcomes (Figure 4—figure supplement 1a). Acini with tumors only differed significantly from those without tumor on the number and proliferation rate of transduced cells in line with the observation that clusters of transduced cells are the primary indicators of tumor outcome. To account for this and possibly other acinus-specific effects, we also evaluated a logistic regression model including the three most important features identified above plus a random effect for acinus of origin and interaction terms for the starting number of normal cells in the acinus. In this model again, the number of oncogene-expressing cells in the cluster appeared as the main contributing factor to tumorigenesis with an odds ratio of 6.7 (Figure 4—figure supplement 1b).

That tumors are mainly initiated by clusters of transduced cells is further illustrated by representative acini shown in Figure 4c; proliferating cells that establish tumor foci at the end of the long-term imaging cluster together at the start of the time lapse imaging and the non-proliferative cells are more sparsely located within the acinus. In line, we found examples of similar sized acini bearing the same number of transduced cells that showed differential cluster expansion depending on the ‘numbers of transduced cells within a cluster’ (Figure 4—figure supplement 2; comparison a to b). Consequently, and again depending on the number of cells within the specific cluster, we observed clusters behaving differently within the same organoid; the organoids shown in Figure 4a, Figure 4c (top and middle panels) and Figure 4—figure supplement 2a, display clusters of single cells that do not expand over time, whereas clusters containing several cells show massive expansion (beyond normal expected cells division rates due to growth of the acinus).

Our results indicate that a proximity-controlled interaction or signaling network between different transformed cells might guide tumor outgrowth in a normal epithelium (Greaves and Maley, 2012; Reid et al., 2010). This might be due to the repressive effect that an intact polarized tissue layer exerts on single early stage cancerous cells (Lee and Vasioukhin, 2008). According to our modeling attempts the simple parameter of ‘number of transduced cells within a cluster’ is enough to impinge on overall epithelial integrity and most likely already accounts for sub-features like ‘distances between transduced cells in a cluster volume’ or ‘number of cell-cell contacts between transduced cells in a cluster volume’, which did not pan out as predictive factors on their own. These observations, however, do not exclude the importance of such traits, which should be explored in more detail for the different situations we report (proliferative versus non proliferative clusters). In addition, the increased probability for tumor outgrowth upon proximity effects of closely located transformed cells, might also be rooted in paracrine effects (e.g. via microRNAs [Kosaka et al., 2012] in the cluster, or hydrogen ion dynamics/pH [Reshkin et al., 2014]). The questions outlined above can now be assessed employing the presented 3D system and longitudinal imaging pipeline together with appropriate fluorescent-tagged reporter proteins.

Studies on loss of important polarity proteins have highlighted their function as non-canonical tumor suppressors in breast tumorigenesis (McCaffrey et al., 2012; Partanen et al., 2012); however, other reports (Macara and McCaffrey, 2013; Xue et al., 2013) cannot confirm these observations which were all obtained from tissue wide transformed model systems, leaving conclusions open for alternative studies. Similarly, it will interesting to trace the effect of prominent tumor factors in breast cancer, such as SCRIBL/MYC (Zhan et al., 2008) or the Hippo pathway (Calses et al., 2019) on overall tissue integrity and promotion of tumor growth in a none tissue wide expression setting. Clearly, to better interrogate the effect of oncogenic drivers on epithelial polarity in more detail, to understand deficiencies in cell-cell interactions towards tumorigenesis and to settle conflicting reports as mentioned above, there is a need for a more detailed analysis, employing a model system that does not show modification of all cells in the tissue.

The interaction of tumor cells with the immediate microenvironment has been subject of extensive studies with regards to immune cells (Binnewies et al., 2018) and other tumor associated cell types (Tabassum and Polyak, 2015); however, the interaction with the normal neighboring cells has not been explored in real time using an organotypic mammalian model system. Rather, questions of cell competition in heterogenous tissues have mainly been addressed in either 2D culture systems or in the Drosophila wing (Tamori and Deng, 2011). Alternatively, organ-specific mammalian cell types that self-organize in a manner similar to in vivo can now be studied in specialized 3D culture conditions in vitro (Lancaster and Knoblich, 2014). Furthermore, these cultures can be used to model disease and serve as an alternative system for drug testing that better recapitulate effects as compared to conventional 2D cell culture (Clevers, 2016; Simian and Bissell, 2017).

Our adaptation of such culture systems, to mark and at the same time enable single cells to express regulatable oncogenes, gave rise to the first stochastic breast tumor model using mammary luminal epithelial acini. Together with a newly established imaging and data processing pipeline, we have developed an integrated approach that allows us to follow cell fates and to interrogate cell-cell interactions of a tumor cell with the normal epithelium. We report that cells within a polarized, differentiated epithelial cell layer are able to establish tumorigenic outgrowth upon stochastic activation of strong oncogenic drivers, supporting the concept of interconversion between differentiated- and stem-like states (Chaffer and Weinberg, 2015; Gupta et al., 2011; Plaks et al., 2015) for tumor initiating cells. However, not all cells expressing the very powerful combination of MYC and Neu oncogenes can expand, demonstrating an inhibitory effect that needs to stem from the normal epithelial microenvironment in the presented system. The ability to distinguish marked tumor cells from the normal epithelium will now allow us to perform single-cell RNA sequencing analysis on select sorted cells. This will help delineate the signaling networks within the immediate tumor microenvironment and further clarify the molecular basis of these distinct processes.

The amenability of 3D multicellular systems to interference with small molecule inhibitors, viral shRNA vectors and genomic editing has the potential to further our understanding of the mechanisms important during tumor initiation. In the future, this will also permit drug screens on the heterogeneous populations of normal and malignant cells, as presented in the described stochastic breast tumor model. The ability to follow tumor regression and establishment of minimal residual disease within a tissue like acinus structure will help distinguish between overall cytotoxic drugs and tumor cell specific inhibitory drugs.

Taken together, we strongly believe that our integration of a true stochastic tumor model with the ability to image single-cell fates will successfully bridge the gap between genetically modified model systems and the clinical situation, helping gain novel insights on breast cancer.

1) Entire image recordings (movies) of time-lapse panels in Figure 3a and 3b (3 Video files in total) have been provided as supplementary movie files. 2) We have uploaded the code for the Feature analysis of the nine acinar features described in Figure 4, as source code file "Feature_Analysis.Rmd". Refer to Supplement file 1 and Online Materials and Methods section for analysis summary. 3) We have uploaded the html file describing the source code as Supplementary file 1. 4) Three. xlsx files with 20 sheets each, one sheet for each acinus analyzed are provided as Supplementary files 2, 3, and 4. These contain the x,y,z coordinates for each cell in the respective acinus at the beginning of the SPIM recording (Supplementary file 2) and at the end (Supplementary file 3). Supplementary File 4 contains the "label" for each transduced cell (corresponding to the labels in Supplement File 2) for the acini at the beginning of the SPIM recording. These. xlsx files were input into the source code to carry out the acinus feature analysis described in Figure 4b and Figure 4 - figure supplement 1. 4) We have deposited the original imaging data for all acini recorded and analyzed (20 mammary acini) at the BioStudies archive at EMBL-EBI (https://www.ebi.ac.uk/biostudies/studies/S-BIAD13). A total of 390-450. h5 image files recorder from 2 channels on the microscope are uploaded for each acini (10 minute time intervals). Raw image data from the microscope was cropped to remove empty pixels, binned in x,y (3,3) and converted to 8-bit images using Big Data Processor Fiji Plug in (http://doi.org/10.5281/zenodo.2574702). This data repository also contains video files generated via Imaris for each acinus, showing fluorescence SPIM miscropscopy data (pre-processed raw files available in respective folders) in 2-color 3D projections (mcherry- magenta; GFP- green) for observing visual phenotypes.

1. 1. Tischer C
   2. Norlin N
   3. Pepperkok R

   (2019) Zenodo

   BigDataProcessor: Fiji plugin for big image data inspection and processing.

   https://doi.org/10.5281/zenodo.2574702
2. 1. Alladin A
   2. Chaible L
   3. Garcia del Valle L
   4. Sabine R
   5. Loeschinger M
   6. Wachsmuth M
   7. Hériché J-K
   8. Tischer C
   9. Jechlinger M

   (2020) BioStudies

   ID S-BIAD13. Tracking the cells of tumor origin in breast organoids by light sheet microscopy - SPIM movie data.

   https://www.ebi.ac.uk/biostudies/studies/S-BIAD13

1. 1. Abe T
   2. Fujimori T

   (2013) Reporter mouse lines for fluorescence imaging

   Development, Growth & Differentiation 55:390–405.

   https://doi.org/10.1111/dgd.12062

   • PubMed
   • Google Scholar
2. 1. Al-Kuraya K
   2. Schraml P
   3. Torhorst J
   4. Tapia C
   5. Zaharieva B
   6. Novotny H
   7. Spichtin H
   8. Maurer R
   9. Mirlacher M
   10. Köchli O
   11. Zuber M
   12. Dieterich H
   13. Mross F
   14. Wilber K
   15. Simon R
   16. Sauter G

   (2004) Prognostic relevance of gene amplifications and coamplifications in breast Cancer

   Cancer Research 64:8534–8540.

   https://doi.org/10.1158/0008-5472.CAN-04-1945

   • PubMed
   • Google Scholar
3. 1. Arteaga CL
   2. Sliwkowski MX
   3. Osborne CK
   4. Perez EA
   5. Puglisi F
   6. Gianni L

   (2012) Treatment of HER2-positive breast Cancer: current status and future perspectives

   Nature Reviews Clinical Oncology 9:16–32.

   https://doi.org/10.1038/nrclinonc.2011.177

   • Google Scholar
4. 1. Binnewies M
   2. Roberts EW
   3. Kersten K
   4. Chan V
   5. Fearon DF
   6. Merad M
   7. Coussens LM
   8. Gabrilovich DI
   9. Ostrand-Rosenberg S
   10. Hedrick CC
   11. Vonderheide RH
   12. Pittet MJ
   13. Jain RK
   14. Zou W
   15. Howcroft TK
   16. Woodhouse EC
   17. Weinberg RA
   18. Krummel MF

   (2018) Understanding the tumor immune microenvironment (TIME) for effective therapy

   Nature Medicine 24:541–550.

   https://doi.org/10.1038/s41591-018-0014-x

   • PubMed
   • Google Scholar
5. Software

   1. Bitplane AG

   (2020) 3/4D Image Visualization and Analysis Software, version v. x64

   Imaris.

   http://bitplane.com
6. 1. Blancato J
   2. Singh B
   3. Liu A
   4. Liao DJ
   5. Dickson RB

   (2004) Correlation of amplification and overexpression of the c-myc oncogene in high-grade breast Cancer: fish, in situ hybridisation and immunohistochemical analyses

   British Journal of Cancer 90:1612–1619.

   https://doi.org/10.1038/sj.bjc.6601703

   • PubMed
   • Google Scholar
7. 1. Bockamp E
   2. Maringer M
   3. Spangenberg C
   4. Fees S
   5. Fraser S
   6. Eshkind L
   7. Oesch F
   8. Zabel B

   (2002) Of mice and models: improved animal models for biomedical research

   Physiological Genomics 11:115–132.

   https://doi.org/10.1152/physiolgenomics.00067.2002

   • PubMed
   • Google Scholar
8. 1. Buckland ST
   2. Burnham KP
   3. Augustin NH

   (1997) Model selection: an integral part of inference

   Biometrics 53:603–618.

   https://doi.org/10.2307/2533961

   • Google Scholar
9. 1. Calcagno V
   2. Mazancourt C

   (2010) Glmulti : an R package for easy automated model selection with (Generalized) Linear models

   Journal of Statistical Software 34:i12.

   https://doi.org/10.18637/jss.v034.i12

   • Google Scholar
10. 1. Calses PC
    2. Crawford JJ
    3. Lill JR
    4. Dey A

    (2019) Hippo pathway in Cancer: aberrant regulation and therapeutic opportunities

    Trends in Cancer 5:297–307.

    https://doi.org/10.1016/j.trecan.2019.04.001

    • PubMed
    • Google Scholar
11. 1. Chaffer CL
    2. Weinberg RA

    (2015) How does multistep tumorigenesis really proceed?

    Cancer Discovery 5:22–24.

    https://doi.org/10.1158/2159-8290.CD-14-0788

    • PubMed
    • Google Scholar
12. 1. Clevers H

    (2016) Modeling development and disease with organoids

    Cell 165:1586–1597.

    https://doi.org/10.1016/j.cell.2016.05.082

    • PubMed
    • Google Scholar
13. 1. D'Cruz CM
    2. Gunther EJ
    3. Boxer RB
    4. Hartman JL
    5. Sintasath L
    6. Moody SE
    7. Cox JD
    8. Ha SI
    9. Belka GK
    10. Golant A
    11. Cardiff RD
    12. Chodosh LA

    (2001) c-MYC induces mammary tumorigenesis by means of a preferred pathway involving spontaneous Kras2 mutations

    Nature Medicine 7:235–239.

    https://doi.org/10.1038/84691

    • PubMed
    • Google Scholar
14. 1. Dekkers JF
    2. Berkers G
    3. Kruisselbrink E
    4. Vonk A
    5. de Jonge HR
    6. Janssens HM
    7. Bronsveld I
    8. van de Graaf EA
    9. Nieuwenhuis EES
    10. Houwen RHJ
    11. Vleggaar FP
    12. Escher JC
    13. de Rijke YB
    14. Majoor CJ
    15. Heijerman HGM
    16. de Winter–de Groot KM
    17. Clevers H
    18. van der Ent CK
    19. Beekman JM

    (2016) Characterizing responses to CFTR-modulating drugs using rectal organoids derived from subjects with cystic fibrosis

    Science Translational Medicine 8:ra344–ra384.

    https://doi.org/10.1126/scitranslmed.aad8278

    • Google Scholar
15. 1. Drost J
    2. van Jaarsveld RH
    3. Ponsioen B
    4. Zimberlin C
    5. van Boxtel R
    6. Buijs A
    7. Sachs N
    8. Overmeer RM
    9. Offerhaus GJ
    10. Begthel H
    11. Korving J
    12. van de Wetering M
    13. Schwank G
    14. Logtenberg M
    15. Cuppen E
    16. Snippert HJ
    17. Medema JP
    18. Kops GJPL
    19. Clevers H

    (2015) Sequential Cancer mutations in cultured human intestinal stem cells

    Nature 521:43–47.

    https://doi.org/10.1038/nature14415

    • Google Scholar
16. 1. Fry EA
    2. Taneja P
    3. Inoue K

    (2016) Clinical applications of mouse models for breast Cancer engaging HER2/neu

    Integrative Cancer Science and Therapeutics 3:593–603.

    https://doi.org/10.15761/ICST.1000210

    • PubMed
    • Google Scholar
17. 1. Greaves M
    2. Maley CC

    (2012) Clonal evolution in cancer

    Nature 481:306–313.

    https://doi.org/10.1038/nature10762

    • PubMed
    • Google Scholar
18. 1. Gupta PB
    2. Fillmore CM
    3. Jiang G
    4. Shapira SD
    5. Tao K
    6. Kuperwasser C
    7. Lander ES

    (2011) Stochastic state transitions give rise to phenotypic equilibrium in populations of Cancer cells

    Cell 146:633–644.

    https://doi.org/10.1016/j.cell.2011.07.026

    • PubMed
    • Google Scholar
19. 1. Havas KM
    2. Milchevskaya V
    3. Radic K
    4. Alladin A
    5. Kafkia E
    6. Garcia M
    7. Stolte J
    8. Klaus B
    9. Rotmensz N
    10. Gibson TJ
    11. Burwinkel B
    12. Schneeweiss A
    13. Pruneri G
    14. Patil KR
    15. Sotillo R
    16. Jechlinger M

    (2017) Metabolic shifts in residual breast Cancer drive tumor recurrence

    Journal of Clinical Investigation 127:2091–2105.

    https://doi.org/10.1172/JCI89914

    • PubMed
    • Google Scholar
20. 1. Held M
    2. Santeramo I
    3. Wilm B
    4. Murray P
    5. Lévy R

    (2018) Ex vivo live cell tracking in kidney organoids using light sheet fluorescence microscopy

    PLOS ONE 13:e0199918.

    https://doi.org/10.1371/journal.pone.0199918

    • PubMed
    • Google Scholar
21. 1. Jechlinger M
    2. Podsypanina K
    3. Varmus H

    (2009) Regulation of transgenes in three-dimensional cultures of primary mouse mammary cells demonstrates oncogene dependence and identifies cells that survive deinduction

    Genes & Development 23:1677–1688.

    https://doi.org/10.1101/gad.1801809

    • PubMed
    • Google Scholar
22. 1. Kosaka N
    2. Iguchi H
    3. Yoshioka Y
    4. Hagiwara K
    5. Takeshita F
    6. Ochiya T

    (2012) Competitive interactions of Cancer cells and normal cells via secretory microRNAs

    Journal of Biological Chemistry 287:1397–1405.

    https://doi.org/10.1074/jbc.M111.288662

    • PubMed
    • Google Scholar
23. 1. Kreso A
    2. Dick JE

    (2014) Evolution of the Cancer stem cell model

    Cell Stem Cell 14:275–291.

    https://doi.org/10.1016/j.stem.2014.02.006

    • PubMed
    • Google Scholar
24. 1. Kutner RH
    2. Zhang XY
    3. Reiser J

    (2009) Production, concentration and titration of pseudotyped HIV-1-based lentiviral vectors

    Nature Protocols 4:495–505.

    https://doi.org/10.1038/nprot.2009.22

    • PubMed
    • Google Scholar
25. 1. Lancaster MA
    2. Knoblich JA

    (2014) Organogenesis in a dish: modeling development and disease using organoid technologies

    Science 345:1247125.

    https://doi.org/10.1126/science.1247125

    • PubMed
    • Google Scholar
26. 1. Langfelder P
    2. Zhang B
    3. Horvath S

    (2008) Defining clusters from a hierarchical cluster tree: the dynamic tree cut package for R

    Bioinformatics 24:719–720.

    https://doi.org/10.1093/bioinformatics/btm563

    • PubMed
    • Google Scholar
27. 1. Lee M
    2. Vasioukhin V

    (2008) Cell polarity and cancer--cell and tissue polarity as a non-canonical tumor suppressor

    Journal of Cell Science 121:1141–1150.

    https://doi.org/10.1242/jcs.016634

    • PubMed
    • Google Scholar
28. 1. Macara IG
    2. McCaffrey L

    (2013) Cell polarity in morphogenesis and metastasis

    Philosophical Transactions of the Royal Society B: Biological Sciences 368:20130012.

    https://doi.org/10.1098/rstb.2013.0012

    • Google Scholar
29. 1. McCaffrey LM
    2. Montalbano J
    3. Mihai C
    4. Macara IG

    (2012) Loss of the Par3 polarity protein promotes breast tumorigenesis and metastasis

    Cancer Cell 22:601–614.

    https://doi.org/10.1016/j.ccr.2012.10.003

    • PubMed
    • Google Scholar
30. 1. McGovern M
    2. Voutev R
    3. Maciejowski J
    4. Corsi AK
    5. Hubbard EJ

    (2009) A "latent niche" mechanism for tumor initiation

    PNAS 106:11617–11622.

    https://doi.org/10.1073/pnas.0903768106

    • PubMed
    • Google Scholar
31. 1. Moody SE
    2. Sarkisian CJ
    3. Hahn KT
    4. Gunther EJ
    5. Pickup S
    6. Dugan KD
    7. Innocent N
    8. Cardiff RD
    9. Schnall MD
    10. Chodosh LA

    (2002) Conditional activation of neu in the mammary epithelium of transgenic mice results in reversible pulmonary metastasis

    Cancer Cell 2:451–461.

    https://doi.org/10.1016/S1535-6108(02)00212-X

    • PubMed
    • Google Scholar
32. 1. Partanen JI
    2. Tervonen TA
    3. Myllynen M
    4. Lind E
    5. Imai M
    6. Katajisto P
    7. Dijkgraaf GJP
    8. Kovanen PE
    9. Makela TP
    10. Werb Z
    11. Klefstrom J

    (2012) Tumor suppressor function of liver kinase B1 (Lkb1) is linked to regulation of epithelial integrity

    PNAS 109:E388–E397.

    https://doi.org/10.1073/pnas.1120421109

    • Google Scholar
33. 1. Plaks V
    2. Kong N
    3. Werb Z

    (2015) The Cancer stem cell niche: how essential is the niche in regulating stemness of tumor cells?

    Cell Stem Cell 16:225–238.

    https://doi.org/10.1016/j.stem.2015.02.015

    • PubMed
    • Google Scholar
34. 1. Podsypanina K
    2. Politi K
    3. Beverly LJ
    4. Varmus HE

    (2008) Oncogene cooperation in tumor maintenance and tumor recurrence in mouse mammary tumors induced by myc and mutant Kras

    PNAS 105:5242–5247.

    https://doi.org/10.1073/pnas.0801197105

    • PubMed
    • Google Scholar
35. 1. Quail DF
    2. Taylor MJ
    3. Postovit LM

    (2012) Microenvironmental regulation of Cancer stem cell phenotypes

    Current Stem Cell Research & Therapy 7:197–216.

    https://doi.org/10.2174/157488812799859838

    • PubMed
    • Google Scholar
36. 1. Reid BJ
    2. Li X
    3. Galipeau PC
    4. Vaughan TL

    (2010) Barrett's oesophagus and oesophageal adenocarcinoma: time for a new synthesis

    Nature Reviews Cancer 10:87–101.

    https://doi.org/10.1038/nrc2773

    • PubMed
    • Google Scholar
37. 1. Reshkin SJ
    2. Greco MR
    3. Cardone RA

    (2014) Role of pHi, and proton transporters in oncogene-driven neoplastic transformation

    Philosophical Transactions of the Royal Society B: Biological Sciences 369:20130100.

    https://doi.org/10.1098/rstb.2013.0100

    • Google Scholar
38. 1. Schindelin J
    2. Arganda-Carreras I
    3. Frise E
    4. Kaynig V
    5. Longair M
    6. Pietzsch T
    7. Preibisch S
    8. Rueden C
    9. Saalfeld S
    10. Schmid B
    11. Tinevez JY
    12. White DJ
    13. Hartenstein V
    14. Eliceiri K
    15. Tomancak P
    16. Cardona A

    (2012) Fiji: an open-source platform for biological-image analysis

    Nature Methods 9:676–682.

    https://doi.org/10.1038/nmeth.2019

    • PubMed
    • Google Scholar
39. 1. Serra D
    2. Mayr U
    3. Boni A
    4. Lukonin I
    5. Rempfler M
    6. Challet Meylan L
    7. Stadler MB
    8. Strnad P
    9. Papasaikas P
    10. Vischi D
    11. Waldt A
    12. Roma G
    13. Liberali P

    (2019) Self-organization and symmetry breaking in intestinal organoid development

    Nature 569:66–72.

    https://doi.org/10.1038/s41586-019-1146-y

    • PubMed
    • Google Scholar
40. 1. Simian M
    2. Bissell MJ

    (2017) Organoids: a historical perspective of thinking in three dimensions

    Journal of Cell Biology 216:31–40.

    https://doi.org/10.1083/jcb.201610056

    • PubMed
    • Google Scholar
41. 1. Tabassum DP
    2. Polyak K

    (2015) Tumorigenesis: it takes a village

    Nature Reviews Cancer 15:473–483.

    https://doi.org/10.1038/nrc3971

    • PubMed
    • Google Scholar
42. 1. Tamori Y
    2. Deng WM

    (2011) Cell competition and its implications for development and Cancer

    Journal of Genetics and Genomics 38:483–495.

    https://doi.org/10.1016/j.jgg.2011.09.006

    • PubMed
    • Google Scholar
43. Software

    1. Tischer C
    2. Norlin N
    3. Pepperkok R

    (2019a) BigDataProcessor; Fiji plugin for visual inspection and processing of big image data, version v0.5.01

    Zenodo.

    http://doi.org/10.5281/zenodo.2574702
44. Software

    1. Tischer C
    2. Pepperkok R

    (2019b) CATS: Fiji plugin for context aware trainable segmentation of big image data, version v0.5.01

    Zenodo.

    http://doi.org/10.5281/zenodo.2574736
45. 1. Verissimo CS
    2. Overmeer RM
    3. Ponsioen B
    4. Drost J
    5. Mertens S
    6. Verlaan-Klink I
    7. Gerwen BV
    8. van der Ven M
    9. Wetering MV
    10. Egan DA
    11. Bernards R
    12. Clevers H
    13. Bos JL
    14. Snippert HJ

    (2016) Targeting mutant RAS in patient-derived colorectal Cancer organoids by combinatorial drug screening

    eLife 5:e18489.

    https://doi.org/10.7554/eLife.18489

    • PubMed
    • Google Scholar
46. 1. Vogelstein B
    2. Fearon ER
    3. Hamilton SR
    4. Kern SE
    5. Preisinger AC
    6. Leppert M
    7. Nakamura Y
    8. White R
    9. Smits AM
    10. Bos JL

    (1988) Genetic alterations during colorectal-tumor development

    New England Journal of Medicine 319:525–532.

    https://doi.org/10.1056/NEJM198809013190901

    • PubMed
    • Google Scholar
47. 1. Williams RT
    2. den Besten W
    3. Sherr CJ

    (2007) Cytokine-dependent imatinib resistance in mouse BCR-ABL+, Arf-null lymphoblastic leukemia

    Genes Development 21:2283–2287.

    https://doi.org/10.1101/gad.1588607

    • Google Scholar
48. 1. Xue B
    2. Krishnamurthy K
    3. Allred DC
    4. Muthuswamy SK

    (2013) Loss of Par3 promotes breast Cancer metastasis by compromising cell-cell cohesion

    Nature Cell Biology 15:189–200.

    https://doi.org/10.1038/ncb2663

    • PubMed
    • Google Scholar
49. 1. Zhan L
    2. Rosenberg A
    3. Bergami KC
    4. Yu M
    5. Xuan Z
    6. Jaffe AB
    7. Allred C
    8. Muthuswamy SK

    (2008) Deregulation of scribble promotes mammary tumorigenesis and reveals a role for cell polarity in carcinoma

    Cell 135:865–878.

    https://doi.org/10.1016/j.cell.2008.09.045

    • PubMed
    • Google Scholar

Author details

1. Ashna Alladin

   Cell Biology and Biophysics Unit, EMBL, Heidelberg, Germany

   Contribution

   Data curation, Formal analysis, Investigation, Visualization, Writing - original draft, Writing - review and editing, Implemented and performed the SPIM imaging and implemented the image analysis workflows, Performed IF analysis

   Contributed equally with

   Lucas Chaible

   Competing interests

   No competing interests declared

2. Lucas Chaible

   Cell Biology and Biophysics Unit, EMBL, Heidelberg, Germany

   Contribution

   Conceptualization, Data curation, Performed the cloning

   Contributed equally with

   Ashna Alladin

   Competing interests

   No competing interests declared

3. Lucia Garcia del Valle

   Cell Biology and Biophysics Unit, EMBL, Heidelberg, Germany

   Contribution

   Data curation, Formal analysis, Performed SPIM imaging and analyzed data

   Competing interests

   No competing interests declared

4. Reither Sabine

   Advanced Light Microscopy Facility, EMBL, Heidelberg, Germany

   Contribution

   Data curation, Validation

   Competing interests

   No competing interests declared

5. Monika Loeschinger

   Luxendo Light-Sheet, Bruker Corporation, Heidelberg, Germany

   Contribution

   Data curation, Validation

   Competing interests

   Monika Loeschinger is employed by Luxendo GmbH, FM BU, Bruker Nano Surfaces, Heidelberg, Germany, the manufacturer of the InVi SPIM light-sheet microscope

6. Malte Wachsmuth

   Luxendo Light-Sheet, Bruker Corporation, Heidelberg, Germany

   Contribution

   Supervision

   Competing interests

   Malte Wachsmuth is employed by Luxendo GmbH, FM BU, Bruker Nano Surfaces, Heidelberg, Germany, the manufacturer of the InVi SPIM light-sheet microscope

7. Jean-Karim Hériché

   Cell Biology and Biophysics Unit, EMBL, Heidelberg, Germany

   Contribution

   Software, Formal analysis, Writing - review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-6867-9425

8. Christian Tischer

   Centre for Bioimage Analysis, EMBL, Heidelberg, Germany

   Contribution

   Resources, Software, Formal analysis, Supervision, Visualization, Writing - review and editing

   Competing interests

   No competing interests declared

9. Martin Jechlinger

   Cell Biology and Biophysics Unit, EMBL, Heidelberg, Germany

   Contribution

   Conceptualization, Supervision, Validation, Writing - original draft, Project administration, Writing - review and editing

   For correspondence

   martin.jechlinger@embl.de

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-3710-4466

• 5,335

  views

• 627

  downloads

• 24

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.