Physical confinement promotes mesenchymal trans-differentiation of invading transformed cells in vivo

Summary Metastasis is tightly linked with poor cancer prognosis, yet it is not clear how transformed cells become invasive carcinomas. We previously discovered that single KRasV12-transformed cells can invade directly from the epithelium by basal cell extrusion. During this process, cells de-differentiate by mechanically pinching off their epithelial determinants, but how they trans-differentiate into a migratory, mesenchymal phenotype is not known. Here, we demonstrate that basally extruded KRasV12-expressing cells become significantly deformed as they invade the zebrafish body. Decreasing the confinement that cells experience after they invade reduces the percentage of KRasV12 cells that trans-differentiate into mesenchymal cell types, while higher confinement increases this percentage. Additionally, increased confinement promotes accumulation of internal masses over time. Altogether, our results suggest that mechanical forces drive not only de-differentiation of KRasV12-transformed epithelial cells as they invade but also their re-differentiation into mesenchymal phenotypes that contribute to distant metastases.


INTRODUCTION
Although metastasis is the predominant cause of mortality in patients with cancer, how tumor cells invade to form metastases is not well understood. For a cancer cell to metastasize, it must first invade from sites where most solid tumors originate, epithelia, and then trans-differentiate to acquire a malignant phenotype. The prevailing metastasis model suggests that as cells accumulate mutations, they first form primary masses from which they later escape by downregulating epithelial-specific genes to undergo an epithelial-to-mesenchymal Transition (EMT) (Thiery, 2002). Transition from epithelia to mesenchymal phenotypes allows cells to dissociate from the primary tumor so that they can invade and colonize distant organs. Because these invading cells have stem-like qualities and markers that allow them to proliferate, survive, and transform into a spectrum of different cell types, the term epithelial-to-mesenchymal plasticity (EMP) has become a better descriptor and more widely used (Bornes et al., 2021).
We have shown that oncogenic KRas mutations that drive aggressive tumors (Aviel-Ronen et al., 2006;Jiang et al., 2009;Neuzillet et al., 2013) induce invasion of transformed epithelial cells by hijacking cell extrusion Slattum et al., 2014), a process that epithelia normally use to promote cell death (Rosenblatt et al., 2001). In epithelial cell extrusion, a basal intercellular actomyosin cable contracts to squeeze one cell apically from the layer to die (Eisenhoffer et al., 2012;Rosenblatt et al., 2001). KRas mutations disrupt apical extrusion, causing cells to accumulate into masses or aberrantly extrude basally, underneath the epithelium Slattum et al., 2014). Importantly, using zebrafish epidermis as a model for simple epithelia where carcinomas form, we previously found that transformed cells basally extrude directly from the epithelium, at sites independent to where they form masses . Basal cell extrusion (BCE) enables KRas-transformed cells to invade, divide, and migrate throughout the zebrafish body. While most invading cells die, they can survive to form large internal masses if the tumor suppressor p53 is absent or mutant, as commonly occurs in aggressive cancers . Importantly for this study, some invading KRas V12 -expressing cells trans-differentiate into mesenchymal phenotypes, with a smaller, but significant fraction adopting a neuronal-like morphology . The mechanisms that regulate differential fates of invaded cells are unknown.
In contrast to previous EMT models, we previously found that de-differentiation of invading transformed epithelial cells occurs suddenly and mechanically as it invades by BCE by pinching off the apical membrane containing E-cadherin and other proteins essential for epithelial identity and function . Whereas all basally extruded cells lack E-cadherin, only a fraction express the mesenchymal markers N-cadherin and snail1b, suggesting a two-step model for EMT whereby cells de-differentiate by BCE and some trans-differentiate into mesenchymal cell types via another mechanism . Given that cells can invade directly from the epithelium, the primary tumor is unlikely to influence the mesenchymal trans-differentiation. However, mechanical stress can induce EMT during embryonic development and tumorigenesis (Fernandez-Sanchez et al., 2015b;Gjorevski et al., 2012), suggesting a potential role for mechanical strain promoting mesenchymal-transformed cell differentiation as cells migrate throughout the tight confines of extracellular matrix and organs within the body. Here, we test the role of mechanical strain on the fate of transformed cells in vivo by altering the microenvironment that cells expressing EGFP-KRas V12 encounter as they invade and migrate throughout the body of zebrafish.

Transformed cells deform following invasion
To investigate the role of the mechanical microenvironment in promoting cell invasion, we mosaically expressed krt4:EGFP-KRas V12 in the outer zebrafish epidermal layer by injecting embryos with a plasmid and a p53 morpholino (MO) to enable invading cell survival. Unless otherwise indicated, all our experiments were quantified at 2 days post-fertilization (dpf). To determine if the tight confinement that invading cells encounter causes mechanical deformation of EGFP-KRas V12 cells, we measured the circularity of their nuclei before and after BCE in embryos expressing mCherry fused to the actin-binding domain of utrophin (Krens et al., 2017) along with H2B-RFP mRNA to visualize nuclei (Megason and Fraser, 2003). We found that the circularity of the nucleus in KRas V12 -expressing cells decreases sharply after BCE ( Figures 1A-1C and Video S1). Furthermore, the nuclear circularity of invaded KRas V12 cells was significantly lower than in those remaining within the embryonic surface, irrespective of their location within the body. (0.83 G 0.01 versus 0.94 G 0.01, p < 0.001, p < 0.01, Figures 1D-1G). The changes in nuclear deformation following invasion suggest that invading cells become confined as they migrate through the dense tissue of the zebrafish body. When cancer cells migrate through confined spaces, nuclear deformation can result in nuclear envelope rupture and DNA damage (Denais et al., 2016;Hatch and Hetzer, 2016;Irianto et al., 2017;Raab et al., 2016). Moreover, compression-derived DNA damage in breast cancer cells results in a snail1-dependent invasive phenotype (Nader et al., 2021). We sought to investigate if nuclear deformation results in DNA damage in invading EGFP-KRas V12 cells by immunostaining with the marker phospho-Histone H2A.X. We found that only 1.3 G 1.3% of KRas V12 -transformed cells in the surface of the embryo had nuclear damage, while 12.23 G 2.4% of invaded KRas V12 -positive cells do so (p < 0.05, Figures 1H-1J). These data suggest that the nuclear deformation cells experience during invasion and migration can damage their nuclei.

Reduced ECM density results in fewer KRas V12 mesenchymal cells
Since we previously found that invading cells could differentiate into mesenchymal and neuronal-like cells following invasion , we next tested if tissue confinement could impact either cell fate. To do so, we tested whether altering the physical environment that EGFP-KRas V12 invading cells encounter affects their trans-differentiation into different cell types. To reduce the extracellular matrix density within embryos (Haeger et al., 2014), we disrupted a key ECM component, laminin, using lamc1 mutants that produce no detectable laminin (Kettleborough et al., 2013) or lamc1 morpholinos that significantly reduce laminin expression (Parsons et al., 2002). Lamc1 mutants have more apoptotic cells in the trunk (Parsons et al., 2002), which could further reduce cell density and confinement. To assess if laminin loss results in reduced overall cell density, we quantified neural tube cell density of control versus lamc1 MO embryos using HRas-EGFP to define cell outlines and DRAQ5 to visualize cell nuclei (Figures 2A and 2B). We found that lamc1 morphants had 28% reduced neural tube cell density compared to controls (p < 0.01, Figure 2C). To investigate if decreasing cell density and confinement affects trans-differentiation, we stained control and lamc1 mutant embryos with the mesenchymal marker N-cadherin ( Figures 2D-2G). The number of invaded cells per embryo did not change between controls (40.2 G 8.2) and lamc1 mutants (30 G 7.1, p > 0.05, Figure 2H). However, 28.8 G 4.0% of invaded KRas V12 /p53MO cells in control embryos express N-cadherin, while only 14.4 G 4.1% of cells invading in lamc1 mutants do so (p < 0.05, Figure 2I). Similarly, lamc1 morphants reduced the number of N-cadherin-positive KRas V12 cells by 77%, compared to controls (p < 0.05, Figures S1A-S1E), despite an 81% increase in the number of invaded cells per embryo in these morphants compared to controls (p < 0.01, Figure S1F). Interestingly, reduced confinement did not significantly alter iScience Article the proportion of invaded KRas V12 /p53MO cells that adopt neuron-like morphologies: 1.1 G 0.7% in lamc1 mutants compared to 2.0 G 1.1% in controls (p > 0.05, Figures 2J-2L). These data suggest that an intact ECM contributes to trans-differentiation of transformed cells into mesenchymal but not neuron-like phenotypes.

Increased confinement promotes trans-differentiation of mesenchymal cell types
Because decreasing the ECM density could affect cell fate by altering signaling via reduced ECM-cell adhesion, we manipulated the mechanical microenvironment independently of ECM composition. To increase physical confinement experienced by KRas V12 -expressing cells following invasion, we embedded embryos from 1 to 2 dpf in 2% agarose to compress them. Agarose-based confinement resulted in significantly smaller embryos, where the total embryo size (area) in 2% agarose (physical confinement) was 0.53 G 0.01 mm 2 ( Figures 3B and 3C) compared to 0.88 G 0.03 mm 2 in 0% agarose (unrestricted growth, p < 0.001, Figures 3A and 3C). Embryos grown in 2% agarose also developed crooked tails ( Figure 3B) and thinner somites along the anterior-posterior axis ( Figures 3D and 3E), suggesting that agarose compressed the tissue during embryonic development. To determine if smaller embryos result from compression or from defects in cell proliferation, we analyzed the cell density of somites in embryos raised in 0 versus 2% agarose expressing HRas-EGFP for cell boundaries and stained with DRAQ5 to visualize cell nuclei. Development in 2% agarose increased the somite cell density by 22%, compared to control embryos (p < 0.05, Figures 3D-3F) but did not affect the number of cells per somite (p > 0.05, Figure 3G), suggesting that agarose causes physical compression, rather than reduced proliferation. To test if increasing confinement affects invading KRas-transformed cell fate, we compared their trans-differentiation into mesenchymal and neuronal-like cells in embryos developing in 0% agarose and in 2% agarose. We found that the number of invaded cells per embryo remained unchanged in 0% and 2% agarose-treated embryos (21.5 G 2.1 and 17.5 G 2.0 respectively, p > 0.05, Figures 3H-3L). Importantly, 26.0 G 2.4% of KRas V12 / p53MO-invaded cells within embryos grown in 0% agarose expressed N-cadherin ( Figures 3H, 3I, and 3M), whereas 38.8 G 3.7% did so in embryos grown in 2% agarose (p < 0.05, Figures 3J, 3K, and 3M). However, no significant differences in the percentage of KRas V12 -positive neuronal-like cells were detected between embryos treated with 0% (6.3 G 1.0%) and 2% agarose (3.9 G 0.9%, p > 0.05, Figures 3N-3P). Together, these results suggest that mechanical compression promotes trans-differentiation of KRas V12invaded cells into mesenchymal but not neuronal cell types.

Altering confinement affects internal mass formation
Embryos injected with KRas V12 in a p53 mutant background develop internalized cell masses by 4.5 dpf, typically within the head . The greater percentage of KRas V12 -expressing mesenchymal cells caused by increased confinement could result in more or bigger masses by 4.5 dpf. However, compression could also cause invading cells to die instead of contributing to these masses. To investigate if mechanical induction of trans-differentiation affects the formation of internal masses, we embedded KRas V12 /p53MO-injected embryos in 0% or 2% agarose from 1 to 2 dpf and fixed them at 4.5 dpf. We found that both types of embryos developed cell masses underneath the E-cadherin-stained epithelium ( Figures 4A and 4B). Importantly, embryos grown in high confinement developed more internal masses than embryos grown in 0% agarose (1.4 G 0.3 vs. 0.6 G 0.2 masses per embryo respectively, p < 0.05, Figure 3. Embryo confinement in agarose promotes KRas V12 cell trans-differentiation into mesenchymal phenotypes (A and B) Example 2 dpf embryos stained for p63 that were grown from 1 to 2 dpf in 0% (A) or 2% agarose (B), with dotted lines outlining embryos. Scale bars, 300 mm.
(G) Number of cells per somite (in xy slice) in embryos grown in 0% agarose (n = 9) or 2% agarose (n = 5). iScience Article Figures 4A-4C). However, the size of the masses remained unaffected in 2% agarose embryos (0.006 G 6.08 3 10 À4 mm 2 ) compared to controls (0.006 G 8.05 3 10 À4 mm 2 , p > 0.05, Figure 4D). These data suggest that mechanically increasing the percentage of KRas V12 -expressing cells that trans-differentiate into mesenchymal phenotypes results in more internal masses by 4.5 dpf. To test if reducing confinement affects the later development of internal masses, we compared the number of cell masses in control KRas V12 /p53MO-injected embryos, and embryos also injected with lamc1 MO. Lamc1 morphants developed fewer internal cell masses than control embryos (0.2 G 0.1 vs. 1.3 G 0.3 masses per embryo, respectively, p < 0.05, Figures 4E-4G). Although there were fewer masses within lamc1 morphants, the sizes of these masses did not vary to those seen in controls (0.006 G 3.2 3 10 À3 mm 2 in lamc1 to 0.005 G 5.81 3 10 À4 mm 2 in controls p > 0.05, Figure 4H). However, the low number of masses in lamc1 morphants makes statistical comparison difficult. Altogether, these results suggest that confinement has a role in the formation of internal cell masses after invasion.

DISCUSSION
Our results suggest that the mechanical deformation that invading transformed cells experience while migrating through confined environments promotes their trans-differentiation into mesenchymal cell types iScience Article and internalized cell masses. Based on our current and previous findings , we suggest that physical forces contribute to cell invasion and formation of distant metastases in two steps: first, apically localized epithelial determinants are pinched off during BCE to de-differentiate epithelial cells (Graphical Abstract, a); second, confinement pressure from cells migrating through tight spaces causes the de-differentiated cells to differentiate into mesenchymal cell that contribute to distant metastases (Graphical Abstract, b).
Our data imply that KRas V12 -invading cells experience DNA damage as they migrate. Previous studies found that DNA damage can cause a partial EMT phenotype in breast cancer cells by the upregulation of snail1 (Nader et al., 2021). Our previous work showed that a fraction of transformed cells express snail1b in our zebrafish system, suggesting a similar mechanism . Altogether, these results suggest that as invading cells migrate, DNA compression and possibly damage could promote trans-differentiation of KRas V12 cells into mesenchymal phenotypes. However, as most cells expressing phospho-Histone H2A.X have invaded, we cannot currently rule out that DNA damage promotes invasion by BCE. Future experiments using live imaging will help to distinguish between both possibilities.
Interestingly, increasing or decreasing the confinement that transformed cells experience only affects their trans-differentiation into mesenchymal but not neural-like cells. It is not clear what impacts the fate of invaded transformed cells with neuronal morphologies. Our previous studies suggested that they can sometimes migrate along normal neurons so local environmental signals may instead contribute to the trans-differentiation of these rarer cell types. Additionally, other internalized KRas V12 /p53cells may differentiate into different unidentified cell types, making it unclear how confinement affects their differentiation.
Furthermore, increasing or decreasing confinement within the body results in more or less internal masses, respectively, by 4.5 dpf, despite the number of invaded cells at 2 dpf remaining the same with each treatment. This indicates that the mechanical forces that cells encounter as they migrate contribute to distant mass formation. Interestingly, increased confinement affects the number, but not size, of cell masses. This could suggest that compression does not affect cell proliferation but does increase oncogenicity. In our system, increasing confinement for just one day resulted in more masses over the long-term, implying that sustained confinement is not necessary for increased KRas V12 cell metastasis. Because most invaded cells die by 4.5 dpf in embryos in a wild-type p53 background , the physical forces cells encounter and nuclear damage may contribute their death. However, when p53 is absent and cell death programs are hindered, these same physical forces could instead promote their trans-differentiation into mesenchymal cell types. In this way, the confined migrating cells encounter could act as a mechanical gymnasium, selecting for survival of the most aggressive, more mesenchymal-like, cells to survive.
Our findings in zebrafish may support a role for compaction forces in tumor cell trans-differentiation and survival into metastases in mammals, as multiple studies have highlighted a role for stiff micro-environments promoting cancer progression (Broders-Bondon et al., 2018;Butcher et al., 2009). For example, at early stages of tumorigenesis, the pressure from tumor hyperproliferation induces expression of the beta-catenin pathway in the surrounding mouse colon, transforming healthy cells into cancer cells (Fernandez-Sanchez et al., 2015a). On the other hand, fibrosis occurring at later stages of disease progression increases tumor environment stiffness that also promotes EMT (Ghajar and Bissell, 2008;Paszek et al., 2005;Weaver et al., 1996Weaver et al., , 1997. Recent discoveries have also shown that the cell nucleus can sense shape changes caused by environmental confinement, causing the cell to adapt its response to surroundings (Lomakin et al., 2020;Venturini et al., 2020). Our findings suggest that confinement pressure directly transforms de-differentiated epithelial cells into mesenchymal phenotypes that promote their survival into distant metastases.

Limitations of the study
While our zebrafish model allows us to follow all invading cells within a single organism, it is important to acknowledge that the signals and forces may be different from those encountered in human cancers. Admittedly, zebrafish embryos may have more embryonic signaling, which could allow for more plasticity than seen in adult cancers. iScience Article of somites and collagen spikes that form fins, forces they would not encounter in human organs. Yet, it is interesting that most (but not all) masses accumulate in the head, which is likely softer tissue (Swift et al., 2013). Additionally, the manipulations we used to alter transformed cell confinement (lamc1 mutants and 2% agarose treatment) could impact cells more broadly, by affecting signaling or developmental morphogenesis, for example. Therefore, we cannot rule out the role that non-mechanical factors might play in the changes to KRas V12 cell mesenchymal trans-differentiation we observe.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

ACKNOWLEDGMENTS
We are grateful to Jon Clarke for sharing the lamc1 morpholino and the Tg(actb1:mCherry-utrCH) and lamc1 sa379 lines with us, to Simon Hughes for sharing the Tg(bAct:hRas-eGFP) line, and to Claudia Linker for the H2B-RFP mRNA. We thank Claudia Linker and the Rosenblatt lab for useful discussions and Rachel Moore for technical help. We are grateful to Alberto Elosegui-Artola and Rachel Moore for comments on the manuscript. T.Z.

DECLARATION OF INTERESTS
The authors declare no competing interests. One or more of the authors of this paper self-identifies as an underrepresented ethnic minority in their field of research or within their geographical location. One or more of the authors of this paper self-identifies as a member of the LGBTQIA+ community.

Agarose confinement
To induce compression, 1 dpf embryos were dechorionated, anaesthetised in MS-222 (Sigma-Aldrich) and embedded in 2% low-melt agarose in E3. Embryos were covered with 0.02% tricaine in PTU-E3, and raised at 28.5 C until 2dpf, when agarose was removed manually by gently scraping it with scalpels. Embryos were immediately fixed using the above protocol. Some of the embryos injected for this experiment were used as controls, covered with 0.02% tricaine in PTU-E3, and raised at 28.5 C until 2 or 4.5 dpf (0% agarose embryos).

QUANTIFICATION AND STATISTICAL ANALYSIS
All our quantitative analysis was performed using SIESTA (Fernandez-Gonzalez and Zallen, 2011;Leung and Fernandez-Gonzalez, 2015), and custom scripts written in MATLAB (Mathworks) using the DIPImage toolbox (TU Delft).

Circularity
To quantify nuclear circularity, we used a mask manually overlaid on the nucleus edge. Nuclear circularity was determined as: where a is the nuclear area and p the perimeter. Circularity is one for circles and lower than one for noncircular shapes. To characterize cells that presented nuclear damage in the surface of the embryo or after invading, embryos injected with EGFP-KRas V12 /p53MO were stained for GFP, phospho-Histone H2A.X as a DNA damage marker and E-cadherin to denote the epithelium to distinguish whether KRas V12 cells had invaded. The number of invaded cells that were both KRas V12 and phospho-Histone H2A.X -positive was quantified. Percentages were calculated as the number of KRas V12 cells expressing phospho-Histone H2A.X, divided by the total number of invaded KRas V12 cells in the surface or inside the embryo.

Cell density
To measure neural tube cell density, z-stacks with a z resolution of 0.5 mm of embryos expressing a membrane marker (HRas-GFP) and a nuclear stain (DRAQ5) were acquired. Then the surface of the neural tube under somites 14-15 was visually detected and a substack was created from 19 mm to 21 mm underneath. A maximum Intensity Projection (MIP) was generated and the area under the somite of interest was delineated by manually overlaying a mask. Finally, the number of nuclei were counted to quantify the cell density as: cell density = number of nuclei area A similar procedure was used to quantify the cell density in the somites. A substack 15 to 20 mm below the surface of somites 14-15 was created and the cell density was quantified as above.

Trans-differentiation quantification
To characterize mesenchymal trans-differentiation levels under different treatments, embryos injected with EGFP-KRas V12 /p53MO were stained for GFP, N-cadherin as a mesenchymal marker and E-cadherin or p63 to delineate the epithelium or the periderm to score only KRas V12 cells that had invaded beneath the periderm. The number of invaded cells that were both KRas V12 and N-cadherin-positive was computed. Neuron-like cells expressing EGFP-KRas V12 were classified based in their morphology, the expression of GFP and their location beneath the periderm. Percentages were calculated as the number of KRas V12 cells expressing N-cadherin or neuron morphology, divided by the total number of invaded KRas V12 cells.

Fish area
To quantify the lateral area of control fish and those embedded in 2% agarose, embryos were stained with p63 to mark the periderm and obtain the embryo outline. Embryos were mounted laterally, and a z-stack of the whole embryo was acquired. MIPs of the resulting images were used to manually segment the fish outline.

Internal cell masses
For the analysis of number and size of internal cell masses by 4.5 dpf, we imaged z-stacks of EGFP-KRas V12 / p53MO-injected embryos, fixed and stained with E-cadherin and GFP. To identify masses, we created binary masks from MIP of the EGFP-KRas V12 signal, by selecting pixels above an intensity threshold. The intensity threshold was the mean image intensity plus four standard deviations. Individual masses were labelled in the binary image and their area was quantified. Labelled objects smaller than 0.0025 mm 2 were ignored for the analysis, to avoid including individual cells. Cells mis-expressing EGFP  were also excluded from the analysis. To ensure that the labelled objects were indeed internal masses we visually confirmed that the EGFP-KRas V12 signal was below the E-cadherin-labelled epithelia in the z-stacks.

STATISTICAL ANALYSIS
To evaluate sample means, we used a non-parametric Mann-Whitney test (Glantz, 2012). To compare more than two groups, we used a Kruskal-Wallis test to reject the null hypothesis, and a Mann-Whitney test with the Holm-Sidak adjustment for pairwise comparisons. Exact value of n and how significance was defined, can be found in the figure legends.