Intravital microscopy of collective invasion plasticity in breast cancer

ABSTRACT Cancer invasion programs are adaptive by switching between metastatic collective and single-cell dissemination; however, current intravital microscopy models for epithelial cancer in mice fail to reliably recreate such invasion plasticity. Using microimplantation of breast cancer spheroids into the murine mammary fat pad and live-cell monitoring, we show microenvironmental conditions and cytoskeletal adaptation during collective to single-cell transition in vivo. E-cadherin-expressing 4T1 and E-cadherin-negative MMT tumors both initiated collective invasion along stromal structures, reflecting invasion patterns in 3D organotypic culture and human primary ductal and lobular carcinoma. Collectively invading cells developed weakly oscillatory actin dynamics, yet provided zones for single-cell transitions with accentuated, more chaotic actin fluctuations. This identifies collective invasion in vivo as a dynamic niche and efficient source for single-cell release.


INTRODUCTION
Progression and fatal outcome of breast cancer disease result from the emergent ability of cancerous cells to invade tissue, cope with complex tissue microenvironments, and adapt their metastatic dissemination programs by switching between collective and individual-cell migration programs (Cheung and Ewald, 2016). Early steps and molecular drivers of single-cell dissemination of epithelial cancers in vivo were identified by intravital microscopy (IVM) in rodent models, including oncogenic mutations in Wnt, EGFR, P53 (also known as Trp53) and TGF-β signaling pathways (Fumagalli et al., 2017;Giampieri et al., 2009;Kedrin et al., 2008). IVM has further revealed how tumors co-evolve with the reactive tumor stroma and undergo anatomic, molecular and functional reprogramming, and the significance of tumor-associated macrophages directing local invasion and systemic dissemination (Friedl and Alexander, 2011;Harney et al., 2015).
In epithelial cancers assessed by histopathological analysis, collective cell patterns are abundant at the invasion front (Bronsert et al., 2014;Cheung et al., 2013;Khalil et al., 2017). Collective invasion occurs in cell groups or strands connected and coordinated by adherens and other cell-cell junctions that mediate multicellular polarity, actomyosin contractility and cell-cell signaling (Friedl and Alexander, 2011). Subsequent to local epithelial cancer invasion, persisting cell-cell interactions can support collective metastasis by tumor cell clusters circulating in peripheral blood and collective organ colonization (Aceto et al., 2014;. However, to date, IVM models of epithelial cancers, including breast cancer and colorectal cancer, have not been able to reliably detect and mechanistically interrogate collective invasion (Fumagalli et al., 2017;Gligorijevic et al., 2014;Kedrin et al., 2007). As a consequence, in vivo insights into collective invasion in epithelial cancers, its guidance by tissue structures, and the mechanisms enabling transitions between collective and single-cell invasion remain lacking.
Here, we applied microsurgical implantation of multicellular breast cancer spheroids into the mammary fat pad, followed by intravital mammary window imaging. From our model, we identified principles of collective invasion, transitions to singlecell dissemination and associated modulation of cytoskeletal states.

Implantation and window-based monitoring of growth and metastasis in mammary tumors
To create a model for monitoring collective invasion of breast cancer cells by intravital microscopy, the mammary imaging model (Kedrin et al., 2008) was adapted for microimplantation of multicellular spheroids at the collagen-containing border of the 4th mammary fat pad (Fig. 1A,B). To maximize throughput, up to 10 spheroids were implanted in the same fat pad (Fig. 1C), mimicking multifocal disease (Hofmeyer et al., 2012). Implanted mouse mammary 4T1 and MMT spheroids contained intercellular junctions including E-cadherin (4T1), β-catenin and p120 catenin (4T1, MMT) ( Fig. S1A-C). The integrity of spheroids, connective and adipose tissue, and vascular networks were preserved after implantation ( Fig. 1B; Fig. S1D), consistent with minimally invasive microsurgery. Multifocal tumors grew exponentially for periods up to 3 weeks ( Fig. 1C; Fig. S1E,F) and developed spontaneous micro-and macrometastasis to the lungs (Fig. 1D,E). In contrast to spheroids, 4T1 cells injected as suspension established bulky tumors without signs of collective invasion (Fig. S1G). Thus, the mammary imaging model recapitulates the growth of primary carcinoma lesions followed by distant metastasis.

Intravital microscopy of collective invasion and individualization
Collective cell invasion in both 4T1 and MMT tumors initiated within 1-2 days after implantation, irrespective of E-cadherin expression ( Fig. 2A; Fig. S1A). 4T1 cells expressing E-cadherin invaded as solid collective strands with detectable leader cells (Fig. 2A,upper row,and B,left). MMT cells which lacked E-cadherin but expressed N-cadherin (Fig. S1A,B), as reported (Firlej et al., 2008), invaded as less tightly organized, but nonetheless collective, strands or networks with multicellular organization and head-to-tail cell alignment (Fig. 2A,lower row,and B,right). Similar collective invasion was obtained when 4T1 and MMT spheroids were implanted into 3D collagen matrix culture, where cells within multicellular invasion strands retained cell-cell junctions enriched in filamentous actin and adherens junction proteins E-cadherin (4T1) and β-catenin (MMT) (Fig. S2A). The ratio of membrane/cytoplasm E-cadherin fluorescence was ∼1.5 along cell-cell junctions in collective strands and decreased to ∼0.5 in detached single cells (Fig. S2B).
Both 4T1 and MMT cells were able to detach and migrate individually with elongated, spindle-shaped morphology ( Fig. 2A,  The main invasion-guiding tissue  structures within the mammary fat pad are represented. An image of the mouse after surgery mounted with a custom-made holder for intravital microscopy is also shown. (B) Z-stack of images showing tissue structure in the mammary fat pad before spheroid implantation (upper row). Z-stack of images of MMT tumor spheroid after implantation into the mammary fat pad (lower row). The image depth is indicated for every image. (C) Multifocal tumor growth monitored by whole-body fluorescence imaging and quantification of multifocal MMT tumor growth. Data represent 6 mice with 7-9 tumors per mouse. Box plots display the median (black line), 25/75 percentiles (boxes) and maximum/minimum (whiskers). P-values were obtained by the Mann-Whitney test. right column) and downregulated E-cadherin at the cell surface (Fig. S2B). Detached single cells were closer to collective strands compared with the spheroid main mass, indicating that single-cell detachment is more likely to occur from collective invasion sites compared with the main tumor mass (Fig. S2C). Individualization was more abundant in MMT compared with 4T1 tumors, which developed more frequent tip cells and detached cells (Fig. 2C). The inverse association between cell individualization and E-cadherin expression in the spheroid implantation model in vivo is consistent with the observed increased single-cell release in 3D organotypic culture of MMT compared with 4T1 spheroids (Fig. S2D), and in patient samples from human lobular compared with ductal breast carcinoma ( Fig. S2E) (Khalil et al., 2017).
Thus, grafted 4T1 and MMT tumors develop predominantly collective invasion of the mammary tissue, and this is consistent with the dominating collective invasion patterns found in human samples of both E-cadherin-positive ductal and E-cadherin-negative lobular breast carcinoma (Bronsert et al., 2014;Cheung et al., 2013;Khalil et al., 2017).

Tissue-guiding structures of mammary carcinoma cells
In the window model, tumor growth and invasion were accompanied by neo-angiogenesis ( Fig. 2A,D) and notable accumulation of fibroblasts at the tumor-stroma interface, similar to human samples (Fig. 3A,B). We mapped the 3D tissue topology next to, and ahead of, the invasion margin to address whether early-onset collective invasion follows microenvironmental structures, a process identified in individually moving breast cancer cells in genetically engineered breast cancer and collectively invading mesenchymal tumors (Gligorijevic et al., 2014;Weigelin et al., 2012). Collective strands, including tip cells, were often aligned parallel to collagen bundles, recapitulating alignment of multicellular strands along stromal collagen in human lesions (Fig. 3C). However, whether earlyonset collective invasion causes remodeling or rather follows preexisting aligned collagen fibrils in vivo is not known. By comparison, individually located 4T1 and MMT cells showed more variable, loosened angle distribution and alignment along collagen structures ( Fig. 3D; Fig. S2F). These data suggest that collective invasion follows more precisely aligned collagen and interfaces, whereas detached single cells are more likely to change direction between guiding tissue structures. Thus, implanted microlesions reproduce collective and individual carcinoma invasion modes and their guidance by stromal structures (Schedin and Keely, 2011).

Adaptive actin dynamics during collective to single-cell transition
Plasticity of cancer cell invasion is associated with an adaptive actin cytoskeleton. In vitro, cortical actin networks define the shape, polarity and force transmission in individually moving cells (Hidalgo-Carcedo et al., 2011), and further stabilize adherens junctions and transmit long-range forces across connected cells during collective migration (Gomez et al., 2011). To address whether actin dynamics in individual and collective invasion differ in vivo, the intensity fluctuations of Lifeact-eGFP in 4T1 tumors transiting from collective to single-cell migration were quantified (Fig. 4A,B). Local actin filament stability, measured by time-delay autocorrelations of Lifeact-eGFP intensity, was high in cells inside (bulk cells) and at the edge (tip cells) of collective invasion zones, but significantly decreased in detached cells (Fig. 4C). This indicates an upregulation of actin dynamics after individualization from the collective invasion zone.
In vitro, amoeboid cells moving along extracellular matrix substrate develop actin waves reaching speeds of ∼10 µm/min (Driscoll et al., 2014), with calculated changes in polymerization on timescales of seconds (Giannone et al., 2007), whereas cell protrusions, such as ruffles or pseudopods, remodel at slower timescales (minutes) (Giannone et al., 2007). To test whether actin dynamics in vivo reach similar kinetics, we probed the intensity variations for 10 and 60 s time delays in collective and single-cell migration in vivo. Detached cells showed a sharp decay of autocorrelation compared with bulk or tip cells (Fig. 4D,E), indicating increased variability of actin dynamics. Likewise, cortical Lifeact-eGFP was focally enriched in individual cells (Fig. 4A) and associated with larger protrusions (>2 µm) and more frequent protrusion growth than in bulk cells, whereas tip cells developed an intermediate protrusion size and frequency (Fig. 4F). In aggregate, these in vivo data suggest an association between increased actin dynamics and decreased precision of tissue guidance after collective to single-cell transition.

DISCUSSION
Although collective invasion patterns are abundantly detected in clinical samples of epithelial cancers, including breast carcinoma (Bronsert et al., 2014;Cheung et al., 2013;Khalil et al., 2017), current intravital microscopy models of breast cancer were insufficient to reliably detect and mechanistically interrogate collective invasion (Giampieri et al., 2009;Gligorijevic et al., 2014;Kedrin et al., 2008). Reasons for this model insufficiency are likely multifactorial, based on (1) lack of cell-cell junctions when tumor cell suspensions are implanted; (2) excessive growth, which might cause circular tissue compression and disable multicellular invasion; and (3) a bias in genetic mouse models, which disable or minimize cadherin-based cell-cell junctions and possibly favor individual cell, but not collective, behaviors. To overcome this problem, we used  (1), tip (2) and bulk (3) cells. Each group of cells exhibits similar background behavior despite differences in actin intensity variability, with detached and tip cells experiencing a higher frequency of large protrusion events than bulk cells. This trend is further indicated in the box plots (E) of the distribution of the 50th and 95th percentiles of boundary motion, albeit no significant differences amongst means were found. (F) The percentages of motion above 2 µm for the indicated cell subsets indicate a significant difference in behavior between detached and bulk, and tip and bulk, cells. Box plots in C, E and F show the median (black line), 25/75 percentiles (boxes), maximum/minimum (whiskers) and outlier (red cross) from 5-6 cells per condition from 3 independent experiments. P-values were obtained by Bonferroni corrected one-way ANOVA and the Tukey-Kramer multiple comparison test. Scale bars: 50 μm (A, overviews), 10 μm (A, areas from dashed line boxes shown magnified on the right).
orthotopic microimplantation of multicellular spheroids of murine breast cancer cells into the mammary fat pad/stroma interface, and monitoredby intravital microscopyreliable collective invasion followed by transition to individual cell detachment in vivo.
Using spheroid microimplantation at the edge of the mammary fat pad, this intravital microscopy approach captures critical early steps of collective invasion and cytoskeletal plasticity in breast cancer tumors. With transition from collective to single-cell migration, individually disseminating cells gain oscillatory behavior of the actin cytoskeleton and abandon strict alignment along collagen fibers. Such collective to single-cell transitions arguably broaden a tumor cell's ability to cope with different tissue topologies and escape from the primary site (Friedl and Alexander, 2011).
In both E-cadherin-positive and E-cadherin-negative cells, implanted spheroids support tumor-like topology and cell-cell cohesion for emerging collective invasion followed by gradual individualization. E-cadherin is considered an important gatekeeper of epithelial functions. Because single-cell events are enhanced in epithelial tumors when E-cadherin is downregulated (Khalil et al., 2017), E-cadherin was initially considered to counteract local tissue invasion and metastatic spread (De Craene and Berx, 2013). However, in clinical breast cancer, the E-cadherin status does not predict metastatic outcome and prognosis (Khalil et al., 2017;Narendra et al., 2015). Similar to these clinical data, E-cadherinnegative MMT cell tumors, albeit developing more frequent cell individualization compared with E-cadherin-expressing 4T1 tumors, did not give rise to increased rates of distant metastasis. Beyond mechanical cell-cell cohesion, other pathways might control the metastatic outcome in MMT and 4T1 cells, including oncogenic signaling, sensitivity to growth factors and mechanoregulation (Friedl and Alexander, 2011;Giampieri et al., 2009;Harney et al., 2015).
Our data further reveal plasticity of the actin cytoskeleton during collective to single-cell transition in vivo, as predicted from in vitro models and in silico (Hidalgo-Carcedo et al., 2011;Te Boekhorst et al., 2016). A relatively quiescent actin cytoskeleton in the collective invasion zone possibly fulfills a dual function in (1) controlling both the stability and turnover of cell-cell junctions, and (2) providing a starting point for actin-rich protrusions at the invasion front. Leader cells which extended outwards to the stroma still retained overall quiescent actin dynamics, comparable to those of bulk cells, but developed large protrusions at rates comparable to those in individual cells. Likely as a consequence of detachment, actin dynamics in individually moving cells unleash oscillations, indicating that the actin cytoskeleton in invading tumor cells is an excitable system with dynamic focalization and protrusion bursts, not unlike traveling actin waves in moving Dictyostelium discoideum amoebae and fibroblasts migrating on 2D substrate in vitro (Driscoll et al., 2012;Giannone et al., 2007). Combined with fate tracing of distinct invasion modes, we consider spheroid microimplantation as a potentially suitable model to barcode invasion mechanisms and stromal niches in primary lesions, and derive their relevance for distant metastasis formation and resistance to therapy (Brighton et al., 2018;Hirata et al., 2015).
When benchmarking in vitro and in vivo analyses of cancer cell invasion, the spheroid microimplantation strategy closes a 'validation gap', as it reflects invasion types and plasticity derived from both 3D organotypic culture and clinical samples (Bronsert et al., 2014;Khalil et al., 2017). Multiple spheroids can be implanted in the same mouse to increase the surface area of the tumor-stroma interface and collect larger data sets. Multiple microlesions further increase the tumor mass and enable spontaneous distant metastasis from these lesions, and this reduces animal consumption and improves data quality from the same mouse. Because intact spheroids are implanted, the strategy will be amenable for epithelial cancer organoids or dissected tumor microtissues from patients to create mouse avatars for personalized medicine.

Spheroid invasion into 3D collagen lattices
Spheroids were incorporated into type I collagen from rat-tail tendon (BD Biosciences; final concentration 4 mg/ml) prior to polymerization at pH 7.4 according to the manufacturer's protocol and overlaid with complete growth medium prior to invasion culture (48 h).

Primary breast cancer samples
Tissue samples from 5 invasive ductal carcinoma (IDC) and 3 invasive lobular carcinoma (ILC) patients were obtained after surgical removal of primary breast cancer (Department of Pathology, Radboudumc, Nijmegen, The Netherlands). Tumor samples were encrypted and analyzed in an anonymized manner, as approved by the Institutional Review Board and according to national law and in compliance with the Declaration of Helsinki (Nagelkerke et al., 2011). According to Dutch legislation, no approval from a research ethics committee was required for this study, as coded tissue obtained from routine diagnostic workflow was used and the included patients were not affected by the study. Anonymous or coded use of redundant tissue for research purposes is part of the standard treatment agreement with patients in Radboudumc Hospital, from which patients may opt out. None of the included patients submitted an objection against use of residual material. Patient material was used in a manner compliant with the Declaration of Helsinki.

In vivo spheroid implantation and mammary window imaging
Balb/C (for 4T1 tumors) and BALB/C-nu/nu (CAnN.Cg-Foxn1nu/Crl; for MMT tumors) female mice (6-8 weeks old) were obtained from Charles River Laboratories. Animal procedures were approved by the Animal Ethics Committee of Radboud University, Nijmegen (RU-DEC 2012-129) and performed according to the guidelines of the Dutch Act on Animal Experimentation and the European FELASA protocol.
For implantation of the mammary window, the 4th mammary fat pad in anesthetized mice (1-2% isoflurane in oxygen) was exposed by surgical incision using microsurgery under a Leica MZFLIII microscope. A superficial microchannel was created within the fat pad using a 30 G syringe needle (BD Medical) into which an individual spheroid in a microdrop (2 µl in PBS) was implanted using a gel-loading pipet tip (0.3 mm inner diameter, BIOplastics). After implantation of up to 10 spheroids in the same tissue area, a mammary imaging window was inserted, affixed to the skin by gentle tightening of a purse-string suture around the skin rim, and closed with a cover glass and spring ring, as described (Alieva et al., 2014). Multifocal tumor development was monitored by whole-body fluorescence imaging (FluorVivo100, INDEC BioSystems) and quantified as total tumor area in the mammary window from 2D projections of the DsRed2 and H2B-eGFP signal.

Spontaneous metastasis assay
4T1 or MMT tumors were grown by implanting 10 spheroids (1000 cells each) into the same mammary fat pad within the mammary imaging window. Two to three weeks after spheroid implantation, the mice were sacrificed and primary tumors and lungs were harvested, snap-frozen in liquid nitrogen (MMT) or fixed in buffered formalin (4%, 24 h) followed by embedding in paraffin (4T1). The absolute number of metastatic events of dual-color MMT cells was counted by sectioning whole lungs into 14 µm slices, with intervals of 70 µm, followed by staining with DAPI and whole-field fluorescence scanning (Pannoramic 250 Flash II) of DAPI, eGFP and DsRed2. For mice bearing 4T1 tumors, which showed less reliable fluorescent protein expression after 3 weeks of growth in vivo, tumor cells were identified by staining for cytokeratin 8 and counted by sectioning whole lungs into 5 µm slices, with intervals of 70 µm, followed by whole-slice scanning (Pannoramic 250 Flash II). Normal lung cells (i.e. pneumocytes) were also detected by cytokeratin 8 staining, and were excluded from quantification based on nonclustered cell morphology, position in alveoli, and smaller size and regular shape of the nucleus compared with nuclei of 4T1 cancer cells. Metastatic load was calculated as the average number of metastatic cell clusters per tissue slice from 74-85 (4T1) and 85-124 (MMT) slices per lung. As a minimum size for inclusion in the analysis, ≥3 4T1 and MMT cells were considered as a metastatic cluster.

Image processing and analysis
Images were obtained using a z-step size of 5 µm, represented as a single slice, maximal projection and 3D reconstruction using Fiji (1.49s, National Institutes of Health). 3D stacks were stitched (mosaic plugin), drift corrected during time-lapse recording (StackReg plugin) and manually segmented to calculate the tumor area. The area of blood vessels was obtained by automated single-channel thresholding of the dextran channel (Alexa Fluor 750) representing perfused blood vessels and normalized to the tumor area. Collagen fiber alignment relative to invading collective strands or individual cells was quantified as the angle between both respective length axes.

Analysis of actin dynamics
Actin dynamics were measured as fluctuations of Lifeact-eGFP fluorescence intensity in space and time from single imaging planes through the center of the cell body without signs of drift. Using a custom semiautomatic MATLAB code (MathWorks), the cell bodies were segmented, and the boundary motion and variability of cortical actin was calculated based on source images with 3.02 µm or 1.45 µm pixel resolution and adjusted signal-to-noise ratio, as described (Driscoll et al., 2014). The cell boundary was defined with subpixel accuracy using an automatic snake algorithm with 200 boundary points. The edge layer of the cell body was defined by subtracting an eroded binary mask of the cell body from a full mask of the cell body. Boundary motion over time was detected by performing a least squares fitting to match boundary points of subsequent frames. The variation of cortical Lifeact-eGFP was calculated for regions near each boundary point along the cell boundary. The boundary length per region was 1.73 µm, which was larger than the average fluctuations in measured boundary position and thus sufficient to capture cortical actin with accuracy. The mean intensity per boundary region over time was plotted as a kymograph and used for calculating the time-delay autocorrelation, defined as: c n ðDtÞ ¼ P T t¼0 ½I n ðt þ DtÞ À I:½ I n ðtÞ À I ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi P T t¼0 ½ I n ðt þ DtÞ À I 2 q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi P T t¼0 ½ I n ðtÞ À I 2 q : In this equation, n represents the boundary index and I an intensity offset to subtract the mean background fluorescence within each cell. The resulting time-delay autocorrelation functions were averaged for given time delays (Δt). Intensity variability was calculated as 1 minus the time-delay autocorrelation represented by 1-c n (Δt). Because a correlation of 1 indicates complete stability of cortical actin with time, this parameter corresponds to the time variability of cortical actin.

Statistics
Statistical analysis was performed with GraphPad software v.5, using the nonparametric Mann-Whitney test, Bonferroni-corrected one-way ANOVA and the Tukey-Kramer test for multiple comparisons. P-values less than 0.05 were considered significant.