Contents lists available at ScienceDirect

Tissue and Cell

journal homepage: www.elsevier.com/locate/tice

Culturing substrates influence the morphological, mechanical andbiochemical features of lung adenocarcinoma cells cultured in 2D or 3D

Adriele Prina-Melloa,b,c,1, Namrata Jaina, Baiyun Liud, Jason I. Kilpatricke, Melissa A. Tuttyc,Alan P. Bella,f, Suzanne P Jarvisc,d, Yuri Volkova,b,c, Dania Moviab,c,⁎,1

a CRANN Institute and AMBER Centre, Trinity College Dublin, Irelandb Laboratory for Biological Characterization of Advanced Materials (LBCAM), Trinity Translational Medicine Institute (TTMI), Trinity College Dublin, Irelandc Department of Clinical Medicine, School of Medicine, Trinity College Dublin, Irelandd School of Physics, University College Dublin, Irelande Conway Institute of Biomedical and Biomolecular Research, University College Dublin, IrelandfAdvanced Microscopy Laboratory (AML), Trinity College Dublin, Ireland

A R T I C L E I N F O

Keywords:3D cell culturesNon-small-cell lung cancerMechanical propertiesHydrogel-based cell culturesMatrigelPuraMatrix™

A B S T R A C T

Alternative models such as three-dimensional (3D) cell cultures represent a distinct milestone towards capturingthe realities of cancer biology in vitro and reduce animal experimentation in the preclinical stage of drug dis-covery. Significant work remains to be done to understand how substrates used in in vitro alternatives influencecancer cells phenotype and drug efficacy responses, so that to accurately link such models to specific in vivodisease scenarios.

Our study describes how the morphological, mechanical and biochemical properties of adenocarcinoma(A549) cells change in response to a 3D environment and varying substrates. Confocal Laser Scanning (LSCM),He-Ion (HIM) and Atomic Force (AFM) microscopies, supported by ELISA and Western blotting, were used. Thesetechniques enabled us to evaluate the shape, cytoskeletal organization, roughness, stiffness and biochemicalsignatures of cells grown within soft 3D matrices (PuraMatrix™ and Matrigel™), and to compare them to those ofcells cultured on two-dimensional glass substrates. Cell cultures are also characterized for their biological re-sponse to docetaxel, a taxane-type drug used in Non-Small-Cell Lung Cancer (NSCLC) treatment. Our results offeran advanced biophysical insight into the properties and potential application of 3D cultures of A549 cells as invitro alternatives in lung cancer research.

1. Introduction

Lung cancer is one of the leading causes of cancer deaths worldwide(Ferlay et al., 2010), and it has one of the highest predicted incidencerates (Malvezzi et al., 2015). Despite the large research and economicinvestments, the survival rate of lung cancer patients remains poor dueto the development of drug resistances (Tsvetkova and Goss, 2012).Consequently, the development of highly effective treatments againstlung cancer remains a vital area of cancer research. In the last years,however, around 95% of new anticancer drugs against lung cancereventually failed in clinical trial (Arrowsmith and Miller, 2013;Hakanson et al., 2014; Kola and Landis, 2004; Landry et al., 2013). Oneof the major reasons for failure, accounting for around 60% of failedclinical trials, is the lack of efficacy in humans despite robust

indications of activity in preclinical models (Arrowsmith, 2011; Hayet al., 2014; Kamb, 2005). Although the reasons for this poor clinicaltranslation are multiple, it has been suggested that major contributorsto such high failure rate may be attributed to the preclinical testingmodels used, which can indeed generate false-positive results in theevaluation of drug efficacy (Arrowsmith and Miller, 2013; Cook et al.,2014; Kamb, 2005; Smalley et al., 2006; Unger et al., 2014).

In conjunction to in vitro two-dimensional (2D) cell cultures, ge-netically engineered mice, ectopic/orthotopic xenografts, and primaryhuman tumourgrafts implanted into immunodeficient mice are the mostcommon preclinical models used in drug development. An importantfeature of all tissues is that cells grow in three dimensions (Abbott,2003; Smalley et al., 2006). Solid tumours are not an exception, andthey are now recognized as complex tissues able to constantly instruct

https://doi.org/10.1016/j.tice.2017.11.003Received 15 September 2017; Received in revised form 31 October 2017; Accepted 26 November 2017

⁎ Corresponding author at: Laboratory for Biological Characterization of Advanced Materials (LBCAM) (Lab 0.74), Trinity Translational Medicine Institute (TTMI), Trinity Centre forHealth Sciences, James’s Street, Dublin 8, Dublin, Ireland.

1 These authors contributed equally to the work.E-mail address: dmovia@tcd.ie (D. Movia).

Tissue and Cell 50 (2018) 15–30

Available online 01 December 20170040-8166/ © 2017 Elsevier Ltd. All rights reserved.

T

and react to chemical and mechanical signals to/from neighbouringcells and the microenvironment (Egeblad et al., 2010). In vitro 2D cellmodels represent therefore simplified microenvironments mimickingthe tumour tissue (Gillet et al., 2013). They indeed impose limitationsin reproducing the cellular behaviour (e.g., division, proliferation, mi-gration, differentiation) of cancer cells (Discher et al., 2005), as well asin reproducing the physical barriers hindering drug permeation into thetumour tissue (Poondru et al., 2002). It is believed that such limitationsnegatively impact upon the relevance of the efficacy data on novelchemotherapeutic agents generated during the preclinical studies(Landry et al., 2013; Rehfeldt et al., 2007), and are also responsible forthe poor in vitro-to-in vivo correlation of drug activities (Poondru et al.,2002). On the other hand, animal models, although indispensable in thediscovery and development process of new cancer drugs, have variouslimitations as preclinical cancer models (Aparicio et al., 2015; Maket al., 2014; Rangarajan and Weinberg, 2003; Ruggeri et al., 2014).Various systematic reviews describe in fact the limitations of animalresearch in the efficacy assessments of new drug candidates (Aparicioet al., 2015; Hackam, 2007; Hartung, 2008; Mak et al., 2014; Olsonet al., 2000; Pound et al., 2004; van der Worp et al., 2010). Thus, al-ternative methods of producing pertinent representations of humansolid tumours must be implemented in the preclinical phase to reliablytest the response of lung cancer cells to specific drugs. Considering thisand in agreement with the “3Rs (Reduction, Refinement and Replace-ment) principle” in animal research, alternative three-dimensional (3D)in vitro models and test platforms have been developed for applicationin lung cancer drug discovery (Nyga et al., 2011; Vinci et al., 2012).

3D culture models can more accurately reflect the complex inter-actions of cancer cells with the surrounding microenvironment, notonly on a biochemical and biomechanical level but also on the level ofgene and protein expression (Friedrich et al., 2009). The pioneeringwork by Mina Bissell's group was central in providing the early ex-perimental evidence that moving cell cultures into 3D systems canmake a difference in cancer research (Weaver et al., 1997). Thus, 3Dcell models have permeated into such research field (Hutmacher,2010), producing a clear shift in preclinical studies towards 3D alter-native systems. Since then, in vitro 3D culture models have been de-veloped from various human tumour cells (Yamada and Cukierman,2007). Furthermore, the continuous progress in tissue engineering(Hakanson et al., 2014), including the development of various 3Dscaffolds, hydrogels and bioreactor systems, has also contributed to theimproved the diversity, fidelity and capacity of such models (Achilliet al., 2012; Friedrich et al., 2007; Stratmann et al., 2014). They havebecome a prevailing alternative to animal experimentation (Kimlinet al., 2013) in the study of: (i) the micro-environmental regulation oftumour cell physiology (Kural and Billiar, 2013); (ii) therapeutic pro-blems associated with metabolic and proliferative gradients in the tu-mour tissue (Rodriguez-Enriquez et al., 2008); and (iii) cell–cell andcell–matrix interactions triggering radio-/chemo-resistance of tumours(Friedrich et al., 2007; Shield et al., 2009). For the latter, 3D co-culturetumour models have been developed. These include heterotypic cellularcomponents (e.g. cancer cells and fibroblasts) and play a critical role inrecreating the tumour microenvironment in vitro. The tumour micro-environment greatly impacts the therapeutic efficiency of chemother-apeutic drugs. Co-culture 3D tumour models represent therefore one ofthe most promising in vitro systems for predictive testing of compoundefficacy in oncology (Hirschhaeuser et al., 2010). Accordingly, nu-merous studies have shown that 3D cell models offer the promise ofimproving clinical translation of drugs by allowing a more accurateexamination of drug sensitivity and resistance (Godugu et al., 2013;Mehta et al., 2012), whilst reducing the need (or number) of animalstudies. As a result, 3D cell cultures can allow for improved in vitro-to-invivo correlations than conventional 2D cell models (Fitzgerald et al.,2015; Hartung, 2017; Weiswald et al., 2013).

Nevertheless, to fully understand how efficacy data originating from3D cell models could be beneficial to the drug discovery pipeline, it is of

prime interest for the cancer research community to characterize suchmodels. The crucial point is to consistently relate the 3D model ofchoice to specific pathological conditions (Friedl et al., 2012). Realis-tically, each in vitro model reflects a particular aspect of the in vivocondition and the ways that cells adapt to that context. While it ishighly intuitive that 2D cell cultures do not recapitulate the real in vivosettings as well as 3D models, it is not so intuitive to actually definewhich in vitro 3D system is as close as possible to a precise in vivo en-vironment. This is due to the current lack of scientific literature de-scribing the morphological, mechanical and biochemical features ofcancer cells cultured in specific 3D microenvironments.

Considering this, in our study we characterized various 3D in vitromodels of Non-Small-Cell Lung Cancer (NSCLC), the most prevalentform of tumour originating from lung epithelial cells. We examined liveor fixed lung adenocarcinoma (A549) cells cultured in 2D on glass or in3D within two different substrates: Matrigel™ and PuraMatrix™. A549cells were chosen as (i) a physiologically relevant in vitro model ofNSCLC (Caino et al., 2012) and as (ii) established cell line for anticancerdrugs discovery and screening (Gazdar et al., 2010). Matrigel™ andPuraMatrix™ were selected as well-accepted hydrogel scaffolds for thecreation of defined 3D microenvironments (Worthington et al., 2015).Generally, scaffolds used for forming 3D in vitro models are groupedinto two broad categories (Horvath et al., 2016). The first includesbiologically-derived/natural materials or hydrogels such as Matrigel™.The second category corresponds to synthetically-derived matrices (e.g.PuraMatrix™). Hence, we decided to focus our investigation to onesubstrate for each of the two categories of scaffolds just described.Matrigel™ is also widely established in cancer research as cell culturesubstrate (Albini and Noonan, 2010). As compared to 2D glass sub-strates, Matrigel™ and PuraMatrix™ have very low storage moduli,corresponding to a soft gel (Georges and Janmey, 2005). The storagemodulus (elastic contribution) value reported in the literature for Ma-trigel™ is 80 Pa as measured by rheometry, while PuraMatrix™ is lessstiff, with a storage modulus value of 5 Pa (Allen et al., 2011).

In our study, we carefully examined the extent to which, shiftingfrom 2D to 3D cell cultures, lung adenocarcinoma cells changed theirshape, cytoskeleton organization and mechanical properties (quantifiedby AFM as cell stiffness), as well as their biochemical signature: ex-pression of Epithelial-to-Mesenchymal Transition (EMT) markers andinvadopodia, cytokine secretion, and activation of Erk1/2 and PI3K/Akt signalling pathways. Finally, we analysed potential differences inthe response to an anticancer drug used in clinical settings (docetaxel).Therefore, to our knowledge, our study provides one of the first in-depth investigations of the physiological features of 3D in vitro modelsthat relate to cancer development, progression and treatment efficacyin human patients, stimulating the adoption of these models in drugdiscovery as alternatives to animal experimentation.

2. Materials and methods

2.1. Chemicals and materials

Chemicals and solvents were purchased from commercial sources(Sigma-Aldrich, Fisher Scientific, Invitrogen and Calbiochem) and usedas received, unless otherwise specified. Glass-bottom petri-dishes werefrom TED-Pella (TED PELLA, Inc., USA) and other cell-culture plasticproducts were purchased from Nunc (Thermo Fisher, Fisher Scientific,Ireland).

2.2. Cell culture

Human alveolar adenocarcinoma cells (A549 cell line) were ob-tained from the American Tissue Culture Collection (ATCC, LGCStandards, Middlesex, UK) and maintained in Hams F12K media sup-plemented with 2 nM L-glutamine, 1% penicillin/streptomycin and10% foetal bovine serum (FBS) (Gibco, Invitrogen Ltd, VWR

A. Prina-Mello et al. Tissue and Cell 50 (2018) 15–30

16

International, Ireland) at 37 °C and 5% CO2. The A549 cell line wasauthenticated using Short Tandem Repeat (STR) profiling (LGCStandards) showing that our A549 batch is an exact match for the ATCChuman cell line CCL-185 (A549) (with 100% match between the sub-mitted sample and the database profile). Mycoplasm and phenotypicresponses were regularly checked for contamination as part of the la-boratory GLP. The passage number was restricted between five andtwelve: cell morphology and mechanical properties were measured andcompared among A549 cells from the same passage. At 80% con-fluence, A549 cells were detached from T75 flask substrate withTryplE™ (Gibco, Invitrogen, Oregon, USA), centrifuged and diluted insupplemented F12K medium at concentrations appropriate for eachexperimental procedure.

2.3. 2D cell models

A549 cells were counted using a Countess™ Automated Cell Counter(Invitrogen, Oregon, USA) and plated in 4-well Millicell® EZ slide(Millipore™, MA, USA) at a concentration of 106 cells/ml (500 μl/well).Cells were incubated for 4 d at 37 °C (5% CO2) to allow cell attachmentonto the glass substrate and cell growth. Cell medium was regularlychanged every 3 d.

2.4. 3D cell models

A549 cells were seeded onto layers (thickness around 1 mm) ofMatrigel™ Basement Membrane Matrix or PuraMatrix™ PeptideHydrogel (both from BD Biosciences, UK). Matrigel™ is a solubilisedbasement membrane preparation extracted from the Engelbreth-Holm-Swarm (EHS) mouse sarcoma, a tumour rich in extracellular matrixproteins; PuraMatrix™ is a synthetic matrix that is used to create de-fined 3D microenvironments for a variety of cell culture experiments.Briefly, Matrigel™ Basement Membrane Matrix was thawed by sub-merging the vial in ice in a 4 °C refrigerator overnight. The vial wasthen swirled to ensure that the matrix was evenly dispersed and, usingcooled pipette tips and keeping the glass slides on ice, the Matrigel™Basement Membrane Matrix was added to the surface of a 4-wellMillicell® EZ slide following the Thick Gel Method reported by thesupplier. This method is recommended by BD Biosciences to grow cellswithin a 3D matrix. Slides were incubated at 37 °C for 30 min prior cellseeding. For PuraMatrix™, the stock solution (1% w/v) was diluted tothe appropriate working concentrations (0.15%, 0.25% and 0.30%)with sterile DI water. PuraMatrix™ can be used in the undiluted form ata concentration of 1% (w/v) or diluted. In our study, selection ofPuraMatrix™ concentrations was based on the manufacturer’s guide-lines for use. According to the latter, concentrations below 0.5% aresuitable for most applications. Thus, a concentration of 0.30% was se-lected to form the 3D cell models. In addition, concentrations equal to0.25% are recommended for surface plating of adherent cells (such asthe A549 cell line), and a final concentration of 0.15% for cell en-capsulation. Hence, we decided to include also these concentrations inour experimental design. 250 μl of PuraMatrix™ dilutions were thenadded to the surface of a 4-well Millicell® EZ slide and gelation waspromoted by carefully and slowly adding supplemented F12K mediumto each well (500 μl/well). The glass slide was incubated at 37 °C for 1 hto complete the gelation of the hydrogel. After the hydrogel has as-sembled, the medium was changed twice over a period of 1 h to equi-librate the environment to reach physiological pH. A549 cells werecarefully seeded at the top of the Matrigel™ or PuraMatrix™ layers at aconcentration of 106 cells/ml (500 μl/well). Cell cultures were grownfor 4 d and cell medium was changed every 3 d.

2.5. Laser scanning confocal microscopy (LSCM)

Immunostaining and LSCM imaging were performed on the 2D/3Dmodels without the need of cell isolation. After fixation with 4%

paraformaldehyde (PFA) for 10 min at room temperature, cell cultureswere permeabilized with 0.1% Triton X-100 in phosphate bufferedsaline (PBS) for 10 min and stained with Hoechst 33342 for nuclei andOregon Green-phalloidin, Alexa Fluor 488®-phalloidin or rhodaminephalloidin (1:50) (all supplied by Invitrogen, Fisher Scientific, Ireland)for F-actin. The slides were incubated at room temperature for 1 h inthe dark, rinsed with PBS and mounted in transparent mountingmedium (VECTASHIELD, Vector Laboratories Inc., CA, USA) prior toLSCM imaging by a ZEISS 510 Meta confocal microscope equipped witha Zen imaging software (Carl Zeiss, Axiovert, Germany). An overnightblocking step (1% bovine serum albumin (BSA) in PBS) was carried outat 4 °C prior to staining when probing for: cortactin (Cortactin (H222)Mouse mAb; 1:100), E-cadherin (E-Cadherin (4A2) Mouse mAb; 1:200),vimentin (Vimentin (D21H3) Rabbit mAb; 1:100) (all purchased fromCell Signaling Technology Inc., Brennan & Company, Ireland), and fi-bronectin (Fibronectin Rabbit mAb; 1:200) (Abcam, VWR International,Ireland). These primary antibodies were diluted in 0.1% Triton X-100 inPBS, added to the cultures overnight at 4 °C, followed by a washing stepand incubation with the appropriate secondary antibody (Goat anti-Rabbit IgG (H + L) Secondary Antibody Alexa Fluor® 488, or Goat anti-Mouse IgG (H + L) Cross-Adsorbed Secondary Antibody Alexa Fluor®488/568) (supplied by Invitrogen, Fisher Scientific, Ireland).Qualitative LSCM imaging was possible by acquiring a series of z-stackimages. Surface rendering of z-stack images was carried out by theopen-source software BioImageXD (Kankaanpaa et al., 2012) and by theVolocity 3D Image Analysis Software (PerkinElmer Inc., Massachusetts,USA) in order to further elucidate the F-actin organization. ImageJsoftware was used for quantifying the fluorescence intensity of the F-actin stain in a single z section.

2.6. Atomic force microscopy (AFM) – imaging and measurements of fixedspecimens

Cell cultures were fixed with 2.5% glutaraldehyde (GA) for 30 minat 37 °C. The samples were washed with PBS three times and thenstored in PBS at 4 °C. AFM measurements were performed on the in vitromodels without the need of cell isolation, by means of a NTEGRASpectra instrument (NT-MDT, Russia) using liquid cell configurations.Soft cantilevers (MLCT probe, Bruker, USA) with nominal values off0 ≈ 15 kHz, k ≈ 0.02 N/m and tip radius ≈ 20 nm were used to probethe cells. Cantilevers were calibrated using the Sader method (Saderet al., 1999).

Topography images were recorded in tapping mode with a resolu-tion of 512 × 512 pixels and a scan rate of 0.4 Hz. For the 3D cellcultures, topography images were acquired only for the cells located ontop of the hydrogels substrates. Data analysis for AFM imaging andsurface roughness was carried out by NOVA Image Analysis Software(NT-MDT, Russia). Cell roughness data was averaged (ncells = 20) foreach sample.

Force mapping was performed in contact mode at 1 Hz. Deflection-height curves represents the movement of piezo (x-axis) during forcecurves against the deflection of the cantilever (y-axis) (SupplementaryFig. S1). Young’s modulus (stiffness) was determined by fitting withHertz model using Protein unfolding and nano-indentation software“PUNIAS” [http://site.voila.fr/punias]. When analysing the 3D cellcultures, force mapping was carried out on cells located on top of hy-drogels substrates, as well as on cells that were partially embedded intothe matrix. Force spectroscopy data are reported as average(ncurves = 500/sample) ± standard deviation.

2.7. Atomic force microscopy (AFM) – force measurements of livespecimens

An AFM (MFP-3D, Asylum Research, USA) was used to quantify themechanical properties of A549 cells forming 2D/3D cultures.Measurements were carried out at room temperature in pre-warmed

A. Prina-Mello et al. Tissue and Cell 50 (2018) 15–30

17

supplemented Hams F12 K culture media. Soft cantilevers (MLCT,Bruker, USA) with nominal values of f0 ≈ 15 kHz, k≈ 0.02 N/m andtip radius ≈ 20 nm were used to probe the cells. The exact springconstant was determined before each experiment for each lever usingthe Sader method (Sader et al., 1999) in air. Inverse Optical LeverSensitivity (InvOLS) was subsequently measured in pre-warmed cellmedium using a glass substrate as a reference. For 3D cell cultures,force mapping was carried out on cells located on top of hydrogels, aswell as on cells that partially penetrated into the matrix. Force mapsarrays of 6 × 6 points were recorded for each cell analysed in a20 × 20 μm area above the cell cytoplasm (rate = 1 Hz; dis-placement ≈ 5 μm; max force = 2 nN) (Supplementary Fig. S2). ForA549 cells grown on glass, arrays of 16 × 16 points were acquired in a30 × 30 μm area due to the more spread out morphology of the cells.The elastic Young’s modulus E, (reported as mean ± relative error andextrapolated from fitting Log(E) distributions) was determined by fit-ting the force vs indentation curve to the Sneddon Model (Sneddon,1965) using Igor Pro (Wavemetrics, USA) and assumed a conical shapedindenter with α = 21.46, Poisson’s ratio ν= 0.5 and a maximum in-dentation depth δ= 200–300 nm to ensure that data was free from anysubstrate influence. Data that did not conform to the Sneddon Modelwere manually excluded from analysis along with any statistical out-liers.

2.8. He-ion microscopy (HIM)

2D and 3D cell cultures were fixed at room temperature in 2.5% GAin 0.1 M Sørensen’s phosphate buffer (pH 7.3) and rinsed withSørensen’s phosphate buffer. After dehydration in increasing con-centrations of EtOH (from 70% up to 100%), the samples were im-mersed in hexamethyldisilazane (HDMS) 30% in EtOH for 5 min beforebeing transferred to HDMS 60% in EtOH for 5 min. The final wash wascarried out in pure HDMS for 10 min. The samples were air dried andimaged by a Zeiss Orion Plus He-ion microscope (Carl Zeiss,Oberkochen, Germany) using an accelerating voltage of 30 kV. Sampleswere transferred into the chamber, which had undergone plasma cleanovernight prior to loading samples, using a load lock. The workingdistance was 8 mm and a 10 μm beam limiting aperture was used. Theprobe current was between 0.5 and 1.5 pA. Images were acquired bycollecting the secondary electrons emitted by the interaction betweenthe He-ion beam and the specimen with an Everhart-Thornley detector(part of the He-ion microscope system). The image signal was acquiredin a 32- or 64-line integration to each contributing line of the image.

He-ion microscopy (HIM) is a relatively new imaging technique inwhich a focused beam of He+ ions is directed onto the sample surface,which liberates secondary electrons that are collected forming detailedimages of the sample surface topography (Bell, 2009). In addressingtopics of the biological sciences, HIM offers various advantages overconventional SEM approaches, such as a high spatial resolution (withbetter material contrast and improved depth of focus) and the ability toimage uncoated, non-conductive samples without the deposition of ametal (or other conductive) overcoat (Rice et al., 2013), which has beenshown to reduce and/or completely mask cell surface details (Bazouet al., 2011). This opens up a whole new range of surface details inbiological specimens that can be examined rapidly and with less risk ofartefacts. Here, we exploited the high-resolution nature of HIM toimage the shape, membrane texture and membranous projections ofA549 cells cultured in 2D or 3D with unsurpassed image resolution anddetail. The authors recognize that the preparation methods of biologicalspecimens for HIM imaging can also introduce artefacts. To avoid thisand to preserve the cells architecture, a drying method using HMDS wasused. HMDS represents a cost- and time-efficient alternative to criticalpoint drying (CPD) in the preparation of cells for SEM imaging (Braetet al., 1997), and recent studies suggest that HMDS-based methods areoptimal to dehydrate cells on 3D scaffolds for SEM examination (Leeand Chow, 2012). Care was taken during sample preparation to ensure

that artefacts were negligible.

2.9. Cell lysis, SDS-PAGE and western immuno-blotting

Cell cultures grown for 4 d were washed with ice-cold PBS. RIPAbuffer (Santa Cruz Biotechnology Inc., Fannin Limited, Dublin, Ireland)supplemented with sodium orthovanadate, protease inhibitor, andPMSF (Santa Cruz Biotechnology Inc., Fannin Limited, Dublin, Ireland)was used as lysis buffer. For cell cultures grown on glass substrates,500 μl/well of supplemented RIPA buffer was added. 1 ml/well ofCorning™ Cell Recovery Solution (Becton Dickinson, Ireland) was addedto allow recovery of cells cultured into Matrigel™. Cell cultures werethen scraped, transferred into a plastic tube, and placed on ice for 1 h.The solution thus obtained was then centrifuged twice at 5000 rpm for5 min, and supernatant decanted after each centrifugation step. Theprecipitate (containing the recovered cells) was re-suspended in 1 ml ofsupplemented RIPA lysis buffer. Cell cultures grown into PuraMatrix™were recovered with a cell scraper, and the suspensions thus obtainedtransferred to a clean Eppendorf tube. After centrifugation at 5000 rpmfor 5 min, the supernatant was decanted and 300 μl of supplementedRIPA lysis buffer added. Following sonication for 15 min in a sonic bathto favour cell lysis, all lysates were centrifuged for 15 min at15,000 rpm at 4 °C. The protein content of each lysate thus obtainedwas quantified using the Pierce BCA Protein Assay Kit (Product no23225; Thermo Scientific, Fisher Scientific, Ireland), as per manu-facturer’s protocol.

For SDS-PAGE, the appropriate lysates volume was diluted in sup-plemented RIPA lysis buffer and NuPAGE® LDS Sample Buffer supple-mented with NuPAGE® Sample Reducing Agent (both supplied byThermo Scientific, Fisher Scientific, Ireland). Samples were heated for10 min at 70 °C, and resolved on pre-casted NuPAGE® 4–12% Bis–TrisGels (Novex®, Life Technologies, Fisher Scientific, Dublin, Ireland) at200 V. A biotinylated protein ladder (Cell Signaling Technology Inc,Brennan & Company, Ireland) was also resolved within the same gel ascontrol. Resolved proteins and molecular weight marker were electro-phoretically transferred to polyvinylidene difluoride membranes(Immobilion P transfer membrane, Merck Millipore, Ireland) by wettransfer for 2 h at 30 V. Polyvinylidene difluoride membranes were thenblocked in either 5% non-fat dry milk in TBS-T 1 × (0.1% Tween20 inTBS) or 5% BSA TBS-T 1 × for 1 h at RT. TBS 10 × was purchased fromSanta Cruz Biotechnology (Ireland), diluted in DI water or 0.1%Tween20-DI water to obtain TBS 1 × and TBS-T 1 × , respectively.Four wash steps (2 × TBS-T 1 × , 2 × TBS 1× , 5 min each in gentleagitation at RT) were carried out before staining with the primary an-tibody (overnight; 4 °C). A list of the primary antibodies used withinthis study (all purchased from Cell Signaling Technology Inc, Brennan &Company, Ireland) is reported in Table 1. Staining of GADPH or α–tu-bulin proteins was used as loading control.

The membranes were then washed and incubated with their re-spective HRP-linked secondary antibodies (anti-rabbit IgG HRP-linkedantibody or anti-mouse IgG HRP-linked antibody, both from CellSignalling Technology Inc, Brennan & Company) (1 h, RT, with gentleagitation). Anti-biotin, HRP-linked Antibody (Cell SignallingTechnology Inc, Brennan & Company, Ireland) was used to stain theprotein ladder. Probed membranes were then washed, incubated withHRP substrate (Luminata™ Forte Western HRP Substrate, MerckMillipore, Ireland) and protein bands visualised by chemiluminescentdetection on CL-XPosure Film (Thermo Scientific, Fisher Scientific,Ireland). Relative proteins’ expression levels were quantified by ImageJsoftware.

2.10. Quantification of secreted cytokine/chemokines, EGF and TGF-β/TGF-β1

Supernatants were harvested from cell cultures at t = 3 d and t = 4d, combined and tested by Enzyme ImmunoSorbent Assays (ELISAs).

A. Prina-Mello et al. Tissue and Cell 50 (2018) 15–30

18

Supernatants were harvested from two independent experiments andtested in duplicate. The amount of human interleukin-1 beta (IL-1β),chemokine (C-X-C Motif) ligand 1 (CXCL-1), human epidermal growthfactor (EGF), latent transforming growth factor beta (TGF-β), total TGF-β1, and free-active TGF-β1 were measured by the following kits:Human IL-1β ELISA MAX™ Deluxe (BioLegend, Medical Supply Co. Ltd,Ireland), Human CXCL-1/GRO-α DuoSet ELISA (R&D Systems Eire,UK), Human EGF ELISA kit (Sigma Aldrich, Ireland), LEGEND MAX™Human Latent TGF-β ELISA, LEGEND MAX™ Total TGF-β1 ELISA, andLEGEND MAX™ Free Active TGF-β1 ELISA (purchased from BioLegend,Medical Supply Co. Ltd, Ireland). Assays were carried out as for man-ufacturers’ protocols. Optical density of each well at 450 nm was de-termined using an Epoch microplate reader (Biotek, MasonTechnologies, Ireland), and was corrected by subtracting the opticalaberration of the 96-well plastic plate at 570 nm.

The cell culture supernatants were also tested by means of PathScanAntibody Array Kit Th1/Th2/Th17 (Cell Signalling Technology Inc,Brennan & Company, Ireland), to simultaneously detect the secretion ofmultiple extracellular signalling molecules. These were: IL-2, IL-4, IL-5,IL-6, IL-8, IL-10, IL-12 p70, IL-13, IL-17A, TNF-α, GM-CSF and IFN-γ.The assay was carried out as for supplier’s protocol: the array was thenincubated with HRP substrate (Luminata™ Forte Western HRPSubstrate, Merck Millipore, Ireland) and the chemiluminescence signalwas detected by FUSION-FX7, equipped with Capt Advance FX7 soft-ware (Vilber Lourmat, France).

2.11. Biological response to docetaxel

After 4 d, 2D and 3D cell cultures were exposed to 10 increasingconcentrations of docetaxel (0.001, 0.05, 0.1, 0.5, 5, 50, 100, 500 and1000 nM) prepared in fresh media. Untreated cultures were included inthe experimental design and used as negative controls (NT). As positivecontrol (PT), cell cultures were exposed to LDH lysis buffer for 45 min.Following 72 h exposure to docetaxel, supernatants were harvested forevaluating the drug cytotoxicity by LDH Cytotoxicity Assay (ThermoScientific, Fisher Scientific, Ireland), while cell cultures viability wasdetermined by quantifying the ATP content in the cultures (CellTiter-Glo® Luminescent Cell Viability Assay, Promega Corporation, MyBioLtd, Ireland). CellTiter-Glo® Luminescent Cell Viability (ntests ≥ 5;nreplicates for each test: 3) and LDH Cytotoxicity assays (ntests ≥ 3;nreplicates for each test: 3) were carried out as for manufacturers’ pro-tocols. Luminescence and absorbance readings were carried out bymeans of FLx800 and Epoch microplate readers (Biotek, MasonTechnologies, Ireland), respectively.

2.12. Statistical analysis

Unless differently stated in the text, Graph-Pad Prism (Graph-PadSoftware Inc., La Jolla, CA, USA) was used to carry out the statisticalanalysis. A p value<0.05 was considered statistically significant. Thestatistical tests used are specified in the corresponding figure caption.

3. Results

In this study, we characterized various in vitro models of lung ade-nocarcinoma grown in 3D within matrices (Matrigel™ andPuraMatrix™). Once seeded on Matrigel™ or PuraMatrix™, cells of theA549 lung adenocarcinoma line invaded the hydrogels (as schemati-cally shown in Fig. 1a) to form 3D cell cultures. For comparison, A549cells were also cultured on glass substrates, obtaining 2D cell cultures.

3.1. Morphology of 3D in vitro models as compared to 2D cell cultures

3.1.1. Cell shapeIt is well documented in the literature that cell-cell and cell-extra-

cellular microenvironment interactions mediate, among others, shape-induced effects (Buxboim et al., 2010; Pickup et al., 2014; Rehfeldtet al., 2007). Accordingly, our results showed that A549 cells re-modelled their shape when cultured in 3D (Fig. 1b). In detail, LSCManalysis demonstrated that the cellular morphology shifted towardsspheroid geometry in 3D cell models.

3.1.2. Cytoskeleton organization and formation of invadopodiaA549 cells re-organized their cytoskeleton as a consequence of the

3D microenvironment (Fig. 1b). Cytoskeletal F-actin was organized intostress-fibres when A549 cells were cultured in 2D, as shown in thefluorescence intensity profiles reported in Fig. 1b, whereas it was cor-tical when cells were cultured in 3D. Interestingly, cytoskeletal F-actinprojections were visible on the cells’ edges only in 3D cell cultures.

Cancer cell invasion is a dynamic process that involves both al-terations in cell shape and formation of membrane protrusions (in-vadopodia), to facilitate invasive growth and migration (Jaafar et al.,2012; Yilmaz and Christofori, 2009). Our morphological data couldtherefore suggest that A549 cells turned to a phenotype with a higherinvasive potential when shifting from 2D to 3D cell organization. In-deed, the changes evidenced herein in the cytoskeletal organization ofA549 cells could also be solely or partially due to the cell adaptation tothe 3D microenvironment itself, and not correlated to a phenotypictransformation. To elucidate this, we evaluated the expression of cor-tactin (Fig. 2), which is a prominent component and a specific markerof invadopodia (Clark et al., 2007), in 2D and 3D cultures. Invadopodiaare specialized extracellular matrix (ECM)-degrading, actin-rich mem-brane protrusions formed by aggressive (mesenchymal) cancer cells,and are the main responsible for tumour cell invasion. Both 2D and 3Dcultures expressed cortactin (Fig. 2a): the highest expression levels ofthis protein were detected in 3D cell cultures grown in Matrigel™ andPuraMatrix™ 0.15% (Fig. 2b). However, careful analysis of LSCMimages showed that, cortactin expression was mainly cortical in 3D cellcultures, while the staining resulted punctuated and intracellular in 2Dcell cultures (Fig. 2a). Notably, in 3D cultures, cortactin strongly co-localized with the F-actin enriched protrusions present on the cellmembrane (as highlighted by arrows in Fig. 2a). This proved that suchprotrusions could indeed be identified as invadopodia. Cortactin as-sembly seemed to be regulated through RhoA-GTPase in all cell cul-tures, while activation of the Rac-GTPase could not be detected(Fig. 2b). Finally, expression of MT1-MMP, which is known to correlateclosely with cortactin expression levels and invadopodia formation(Clark et al., 2007), was found to be expressed in A549 cells grown in3D in Matrigel™ and PuraMatrix™ 0.15% (Fig. 4c), thus supporting theconclusion that A549 cells formed invadopodia when cultured in 3D.

It should be highlighted that, in some instances, our experimental

Table 1Primary antibodies used in this study, the dilution used and the diluent in which theywere prepared.

Primary antibody Dilution Diluent

phospho-p44/42 MAPK (Erk1/2) (Thr202/Tyr204) Antibody

1:2000 5% BSA in TBS-T 1×

phospho-Akt (Ser473) (D9E) XP® RabbitmAb

p44/42 MAPK (Erk1/2) (137F5) RabbitmAb

1:1000

Akt (pan) (C67E7) Rabbit mAbRhoA (67B9) Rabbit mAbRac1/2/3 AntibodyCdc42 (11A11) Rabbit mAbphospho-mTOR (Ser2448) (D9C2) XP®

Rabbit mAbMT1-MMP (D1E4) Rabbit mAbCortactin (H222) Antibody 1:1000 5% non-fat dry milk in

TBS-T 1×E-Cadherin (4A2) Mouse mAbGADPH (D16H11) XP® Rabbit mAbα–tubulin Antibody

A. Prina-Mello et al. Tissue and Cell 50 (2018) 15–30

19

results displayed differences in the protein expression levels of pairs ofreplicates (Fig. 4b and c). When working with 3D cell cultures, it iscommon to experience variability. This is considered one of the mainissues preventing the successful integration of these models in theroutine preclinical screening of anti-cancer drugs. Though it is difficultto pinpoint the exact causes, a factor that could have contributed todifferences in protein expression in our study was the batch-to-batchvariability in hydrogel composition (Matrigel™) or polymerization(PuraMatrix™). A deeper level of detail could be obtained by studyingthe effect of these parameters on the 3D cultures. A better

understanding on the influences that these parameters could have onthe 3D cultures’ characteristics would be the matter of a further indepth study.

3.1.3. Expression of EMT biomarkersTo further elucidate the changes in the cell invasive potential of

A549 cells as a consequence of the microenvironment, we evaluated theexpression of Epithelial-to-Mesenchymal (EMT) protein markers in 2Dor 3D cultures (Fig. 3).

Loss of E-cadherin is one of the best indicators of EMT transition in

Fig. 1. Cell shape. (a) Schematic of the 2D (glass) and 3D (hydrogels) cell cultures. (b) Schematics of specimens and representative LSCM images of A549 cells cultured on (from top tobottom) glass, Matrigel™ and PuraMatrix™ (at three different concentrations: 0.15%, 0.25% and 0.30%). Projections and rendered reconstructions of series of z-stack images are reported.Cells were stained with Oregon green-phalloidin (F-actin, in green) and Hoechst 33342 (nuclei, in blue). Arrows clearly point out the formation of lamellipodia, filopodia and/orinvadopodia on the cells’ edges in 3D cell cultures. The fluorescence intensity of the F-actin stain across the yellow lines was quantified using ImageJ software and the intensity profile isgraphed. Scale bars: 22 μm (2D cell culture) and 10 μm (3D cell cultures) (63×objective lens). (For interpretation of the references to colour in this figure legend, the reader is referred tothe web version of this article.)

A. Prina-Mello et al. Tissue and Cell 50 (2018) 15–30

20

epithelial cells (Thiery, 2002). Our results showed that E-cadherin wasexpressed only in 3D cultures grown in Matrigel™ and PuraMatrix™0.15% (Fig. 3a). This was confirmed by Western blotting (Supplemen-tary Fig. S3). Not surprisingly, the mesenchymal protein vimentin wasexpressed in all cell cultures (2D and 3D) (Fig. 3b). Vimentin is in fact aprotein normally expressed in cells of mesenchymal origin, such as theA549 cell line. Fibronectin expression was also detected in all cultures,with the highest levels found in 3D cultures grown in Matrigel™(Fig. 3c).

3.2. Topographical characterization of 3D cell cultures

3.2.1. Cell surface roughnessAtomic Force Microscopy (AFM) and He-Ion Microscopy (HIM)

were used to characterize the cell topography and describe in detail thesubtle cell membrane structural changes evidenced by the morpholo-gical analysis (Section 3.1).

Detailed topographical images recorded by AFM (Fig. 4a), showedthat fixed A549 cells cultured on glass (2D) had flattened morphologyand presented actin-enriched regions at cells’ edges with a height ofabout 0.5–0.8 μm. This correlated with a rather smooth cell surface, asdetermined from surface roughness measurements. The round cell ag-gregates formed by A549 cells cultured in Matrigel™ were characterizedby a smooth cell surface, comparable to that of 2D cell cultures. Thisresult agreed with previous scientific literature (Cichon et al., 2012).On the other hand, AFM topographic images recorded an increase insurface roughness for A549 cells cultured in PuraMatrix™, with insig-nificant differences among the three PuraMatrix™ concentrationstested.

The AFM topographical results were confirmed by HIM images(Fig. 4b), which evidenced the presence of unique membrane rufflesand a multitude of filopodia/invadopodia (highlighted by arrows) onthe surface of A549 cells cultured within PuraMatrix™. In this instance,HIM allowed avoiding conventional Scanning Emission Microscopy

Fig. 2. Expression of invadopodia in 2D and 3D cell cultures. (a) Representative LSCM images of A549 cells cultured on (from left to right) glass as 2D culture, or in 3D in Matrigel™ andPuraMatrix™ (at three different concentrations: 0.15%, 0.25% and 0.30%). Projections of series of z-stack images are reported. Cells were stained with Cortactin (H222) Mouse mAbfollowed by Goat anti-Mouse IgG (H + L) Cross-Adsorbed Secondary Antibody Alexa Fluor® 488 (cortactin, in green), Hoechst 33342 (nuclei, in blue), and rhodamine-phalloidin (F-actin,in red). Arrows highlights the co-localization of cortactin in F-actin enriched protrusions of the cell membrane. Scale bars: 20 μm (63 × objective lens). (b,c) Western blot showing theexpression of: (b) cortactin, RhoA-GTPase, Rac1/2/3-GTPase and (c) MT1-MMP in A549 cells cultured in (from left to right): 2D on glass substrates (G), or 3D in Matrigel™ (M) orPuraMatrix™ (PM) at three different concentrations (0.15%, 0.25% and 0.30%). Two individual experiments (n1 and n2) are presented for each culture. The relative expression levels ofcortactin, RhoA-GTPase and MT1-MMP in each sample was calculated as the ratio between the protein band intensity and that of the loading control (GADPH). (For interpretation of thereferences to colour in this figure legend, the reader is referred to the web version of this article.)

A. Prina-Mello et al. Tissue and Cell 50 (2018) 15–30

21

(SEM) sample preparation artefacts associated with metal coatingprocedures. The low sputter yield rate of the helium ions beam fororganic materials (Joens et al., 2013), in addition to the increased HIMresolution versus SEM, the advanced charge reduction and the highcontrast secondary electron emission (Bell, 2009), enabled the repeatedimaging of small and delicate surface features and structure of com-pletely unmodified 3D cell cultures at the nanometre scale with no

discernible damage by the beam. In our study, HIM represented a time-efficient methodological approach that could perfectly match and/orcomplement time-consuming AFM topographical measurements, foridentifying the subtle cell morphological changes induced by the 3Dmicroenvironment. Nevertheless, we recognize that the methodologydescribed here has some limitations, since the AFM/HIM approachtaken allowed recording the cellular morphological properties only of

Fig. 3. Expression of EMT markers in 2D and 3D cell cultures. Representative LSCM images of A549 cells cultured on (from left to right) glass, in Matrigel™ and in PuraMatrix™ (at threedifferent concentrations: 0.15%, 0.25% and 0.30%). Projections of series of z-stack images are reported. Cells were stained with: (a) E-Cadherin (4A2) Mouse mAb followed by Goat anti-Mouse IgG (H + L) Cross-Adsorbed Secondary Antibody Alexa Fluor® 568 (E-cadherin, in red), Hoechst 33342 (nuclei, in blue), and Alexa Fluor 488®-phalloidin (F-actin, in green). (b)Vimentin (D21H3) Rabbit mAb coupled to Goat anti-Rabbit IgG (H + L) Secondary Antibody Alexa Fluor® 488 (vimentin, in green), Hoechst 33342 (nuclei, in blue), and rhodamine-phalloidin (F-actin, in red). (c) Fibronectin Rabbit mAb followed by Goat anti-Mouse IgG (H + L) Cross-Adsorbed Secondary Antibody Alexa Fluor® 568 (fibronectin, in red), Hoechst33342 (nuclei, in blue), and Alexa Fluor 488®-phalloidin (F-actin, in green). Scale bars: 20 μm (63× objective lens). (For interpretation of the references to colour in this figure legend, thereader is referred to the web version of this article.)

A. Prina-Mello et al. Tissue and Cell 50 (2018) 15–30

22

those cells located on top of hydrogels substrates, and not of the entire3D cell population.

3.3. Mechanical properties of 2D and 3D cell cultures

3.3.1. Cell stiffnessCellular stiffness was investigated in both fixed and live specimens

by AFM nanoindentation (Radmacher, 1997; Touhami et al., 2003), tofurther elucidate the effect of the 3D microenvironment on the cellularphenotype and invasive potential of A549 cells. The micro-fabricatedprobe used in AFM offers force sensitivity on the order of 10 pN, toprobe the elasticity of very soft and inhomogeneous specimens (such as3D cell cultures) with a spatial resolution at molecular or atomic scales(Dufrene et al., 2013). To date, AFM nanoindentation studies (Lekka

and Laidler, 2009; Morton and Baker, 2014; Sokolov, 2007) haveshown that cancer cells of various origins are generally characterized bydecreased stiffness (due to changes in the cytoskeletal organization) ascompared to healthy samples (Cross et al., 2007; Cross et al., 2008;Dokukin et al., 2011; Iyer et al., 2009), with sensitivity and specificityup to 99% (Guz et al., 2015). Additionally, AFM data has demonstratedthat changes in cancer cell stiffness could distinguish cancer cells withhigh metastatic potential from less invasive types (Liu et al., 2014;Watanabe et al., 2012; Xu et al., 2012; Zhou et al., 2013). In this study,analysis of force-distance curves allowed us to determine the stiffness(here reported as Young’s modulus) of A549 cells in the different 2D/3Dmicroenvironments. Our results on fixed specimens showed that, cellswere stiffer when cultured on 2D glass substrates compared to 3Dmodels (Fig. 5).

Fig. 4. Cell surface roughness. (a) AFM surface topography and average roughness of fixed A549 cells cultured on (from top to bottom) glass, or within Matrigel™ and PuraMatrix™ (atthree different concentrations: 0.15%, 0.25% and 0.30%). Schematics depict the experimental setup: for schematic purposes, only the interaction of the AFM tip with a surface A549 cell isshown. Arrows point out examples of F-actin-enriched regions in a cell cultured in 2D. Scale bars: 10 μm. Average roughness data are reported as arithmetic mean(nreplicates = 20) ± standard deviation. (b) Representative HIM images of A549 cells cultured in (from top to bottom) 2D on glass, or as 3D cell model growing within Matrigel™ orPuraMatrix™ (at concentration equal to 0.15%, 0.25% or 0.30%). Arrows highlight characteristic filopodia/invadopodia protruding from cells’ edges in 3D cell cultures grown insidePuraMatrix™.

A. Prina-Mello et al. Tissue and Cell 50 (2018) 15–30

23

Since the sample fixation procedures could indeed alter the me-chanical properties of the cells and of the surrounding hydrogel-basedmatrix (Matrigel™ and PuraMatrix™), AFM force-distance curve analysiswas used to also investigate the characteristics of live specimens. Theresulting elastic Young’s modulus (E) of live A549 cells is shown inFig. 6 as histograms of the logarithmic function of E (Log10(E)). For cellsgrown in 2D on a glass surfaces, the histogram of Log10(E) could befitted with a single Gaussian distribution centred at 3.05 ± 0.49 kPa(Fig. 6a). It can be hypothesized that the broad distribution is asso-ciated with the variance in stiffness of the different components of thecells (such as nucleus and a highly-organized cytoskeleton). Log10(E)histograms were narrowly distributed for live A549 cells grown withinMatrigel™, with a maximum peak at around 2.8 kPa (Fig. 6b). In con-trast, cells cultured within PuraMatrix™ 0.15% exhibited a bimodaldistribution for the Log10(E) histogram, with a dominant peak centredat 1.25 ± 0.09 kPa, implying a soft cell response, and a second peakexhibiting a stiffer cell response at 20.80 ± 1.56 kPa (Fig. 6c). Suchbimodal Log10(E) distribution suggested the coexistence of cells withdifferent cytoskeletal organizations (and therefore mechanical proper-ties) within the same specimen. We suggest that the cells’ spatial

localization within the 3D cultures played a major role in defining theLog10(E) distribution. We hypothesise that cells partially embedded intothe hydrogel-based matrix were more rigid as compared to the cellslocated on the surface of the hydrogel substrate, in response to theircloser proximity to the glass substrate. Due to the semi-liquid nature ofPuraMatrix™ at very low concentrations (e.g. 0.15%), such cells might“feel” the glass substrate and respond by adapting their mechanicalproperties. A transition from the bimodal distribution to a normal dis-tribution was in fact evident when shifting to a more viscous (andtherefore stiffer) substrate, such as PuraMatrix™ at higher percentages(0.25% and 0.30%). For A549 cells grown within PuraMatrix™ 0.25%(Fig. 6d), a dominant response occurred at 0.50 ± 0.05 kPa with asmall tail at 3.34 ± 0.12 kPa, whereas cell cultures formed withinPuraMatrix™ 0.30% showed a single peak at 2.44 ± 0.16 kPa (Fig. 6e).On average, live A549 cells grown in 3D were found to be less stiff thanthose cells cultured in 2D on glass (Fig. 6f). This confirmed the AFMnanoindentation results obtained on fixed specimens.

3.4. Biochemical signatures of 3D cell cultures of A549 cells

3.4.1. Cell viabilityWe evaluated the ATP levels in untreated 2D and 3D cell cultures as

an indicator of viability. No significant differences could be detected,with exception of A549 cells cultured on PuraMatrix™ 0.30% thatshowed an increased cell viability (Fig. 7).

3.4.2. Secretion of cytokines/chemokines, EGF and TGFβ-1 and subsequentactivation of specific cellular pathways

Cancer cells are known to increase the secretion of specific factors,including cytokines/chemokines and growth factors (e.g. EGF andTGFβ-1), that play an important role in tumour progression and che-moresistance. In our study, no significant secretion of IL-2, IL-4, IL-5,IL-6, IL-10, IL-12 p70, IL-13, IL-17A, TNF-α, GM-CSF, IFN-γ (Fig. 8a)and IL-1β (Fig. 8b) could be detected in supernatants harvested from 2Dor 3D cell cultures.

On the other hand, IL-8 (also known as CXCL-8) and CXCL-1 (or

Fig. 5. Cell stiffness in fixed specimens. Young’s Modulus (i.e., stiffness) of fixed A549cells cultured (from left to right) in 2D (on glass, G) and in 3D within Matrigel™ (M) orPuraMatrix™ (PM) at three different concentrations (0.15%, 0.25% and 0.30%). For eachsample data are reported as mean (nreplicates = 500)± standard deviation.

Fig. 6. Stiffness of live cells. (a-e) Distribution of the loga-rithmic function of the elastic modulus (Log10(E)) measuredby AFM of live A549 cells grown on (a) glass substrates in 2D(nreplicates = 2135), or in 3D within (b) Matrigel™(nreplicates = 133) and (c–e) PuraMatrix™ (nreplicates = 429,164 and 126 for 0.15%, 0.25% and 0.30%, respectively). TheKruskal-Wallis (KW) test for nonparametric samples (whichdoes not assume any specific distribution) showed that all thedistributions observed were statically independent (P = 0.0).(f) Box-and-whisker plot of Log10(E) of live A549 cells grownon (from left to right) glass (G), Matrigel™ (M) andPuraMatrix™ (PM) at three concentrations (0.15%, 0.25% and0.30%). The median E value (indicated by a blue line) forA549 cells grown in 2D on glass is superimposed with a greydashed horizontal line, and highlights that cells cultured in3D within Matrigel™ or PuraMatrix™ had a lower stiffness.(For interpretation of the references to colour in this figurelegend, the reader is referred to the web version of this ar-ticle.)

A. Prina-Mello et al. Tissue and Cell 50 (2018) 15–30

24

growth-related oncogene-α, GRO-α), were found in the supernatants ofboth 2D and 3D cell cultures (Fig. 8a and c, respectively). Notably, thelevels of IL-8/CXCL-8 detected were higher when A549 cells weregrown in 3D culturing conditions, with the highest amounts found insupernatants harvested from 3D in vitro models grown in Matrigel™ orPuraMatrix™ 0.30% (p < 0.05 as compared to 2D cell cultures). Both2D and 3D cell cultures showed expression of phospho-Akt andphospho-mTOR (Fig. 8d): phosphorylation of such proteins is known tobe triggered by, among other factors, IL-8 through phosphatidylino-sitol-3 kinase (PI3K) activation (Liu et al., 2016; Waugh and Wilson,2008). PI3K activation induces phosphorylation of its substrate, Akt,and subsequently of mTOR: phospho-mTOR is involved in cancer cellssurvival through autophagy inhibition. While no significant differenceswere found in the expression levels of phospho-Akt among cultures, thehighest amounts of phospho-mTOR were detected in 3D cell culturesgrown in Matrigel™. CXCL-1 release to the extracellular space, whichcan ultimately trigger cancer cells’ tumorigenicity (Zhong et al., 2008),is also thought to be regulated through PI3K/Akt pathway activationand subsequent phosphorylation of mTOR (Lo et al., 2014; Shieh et al.,2014). Accordingly, Matrigel™-based 3D cultures showed an increase(although not significant) in CXCL-1 secretion levels (Fig. 8c). IL-8signalling also activates mitogen-activated protein kinase (MAPK) cas-cade, with downstream phosphorylation of Erk1/2 (Liu et al., 2016;Waugh and Wilson, 2008). In our study, the protein bands corre-sponding to phospho-Erk1/2 were clearly detectable in all cultures(Fig. 8f). Interestingly, no EGF could be detected in the supernatants of2D/3D models (Fig. 8e). Activation of Erk1/2 independently from EGFultimately describes a pathway linking IL-8 signalling to the phos-phorylation of such proteins. 2D cell cultures and 3D models grown inMatrigel™ shared similar expression levels of phospho-Erk1/2; whereas,an increased phosphorylation of these proteins was clearly detected in3D cultures grown in PuraMatrix™ (Fig. 8f). Significant differencescould be also found when varying the matrix concentration, from0.15% to 0.30%, with the highest expression levels found in 3D culturesgrown in PuraMatrix™ 0.30%. IL-8/CXCL-8 can also trigger activationof Rho-GTPase, promoting invasion and motility through the poly-merization of cortactin (Liu et al., 2016; Waugh and Wilson, 2008). Insection 3.1.2 we have shown that RhoA-GTPase, as well as cortactin,expression could be detected in all cell cultures (Fig. 2).

Finally, a decrease in free-active TGFβ-1 levels were evidenced insupernatants harvested from 3D cell cultures (Fig. 8g), with the mostsignificant decrease associated to cells cultured in Matrigel™. The se-cretion levels of latent TGFβ and total TGFβ-1 were also quantified as acontrol (Supplementary Fig. S4). Latent TGFβ secretion levels followeda trend similar to that described for free-active TGFβ-1; whereas, thedetectable amounts of total TGFβ-1 were similar in all cell cultures. Thissuggested that the enzymatic activity necessary to activate TGFβ-1varied among 2D and 3D culturing conditions. A549 cells cultured in

2D responded by phosphorylating SMAD2 when stimulated with re-combinant human (rh)-TGFβ-1 (5 ng/ml) for 24 h (Fig. 8h) or as re-sponse to the autogenous levels of TGF-β1 detected in culture after 4 d(Fig. 8i). No activation of the SMAD2 signalling cascade was evidencedin 3D cultures (Fig. 8i).

3.5. Chemoresistance

Data on the ATP levels in untreated (NT) and docetaxel-treated 2D/3D in vitro models showed that a decrease in the cell viability close tothe half-maximal lethal concentration (LC50) could be detected only in3D cell cultures grown in PuraMatrix™ 0.25% and PuraMatrix™ 0.30%for docetaxel concentrations above 50 nM (Fig. 9a). Furthermore,docetaxel-triggered cytotoxicity could be detected in these two culturesat concentrations equal or above 100 nM and 500 nM, respectively(Fig. 9b).

To further validate our chemoresistance results, the drug sensitivityof the A549 cell line batch used in this study was validated against theGDSC (Genomics of Drug Sensitivity in Cancer) database benchmark(Yang et al., 2013). Sub-confluent 2D cultures of A549 cells were ex-posed to docetaxel at its nominal IC50 concentration (1 nM), as reportedby the GDSC database for A549 cells. As predicted during the design ofthe experiment, a significant decrease (p < 0.001) in cell viability wasdetected in sub-confluent 2D cultures as a consequence of exposure todocetaxel (Fig. S5 in the Supporting Material). Consistently with thedata provided by GDSC database, such decrease was equal or close to50% for all drugs (Fig. S5 in the Supporting Material). Hence, wesuggest that the lack of significant cell death in 2D cell cultures shownin Fig. 9 was associated to the long culturing times, which allowed cellproliferation and formation of highly confluent monolayers. This isconsistent with data reported by other research groups, showing thatdrug efficacy is reduced when cancer cells are cultured at high density(Lieberman et al., 2001).

4. Discussion

It is well recognized that cancer cells respond to 3D architecture andstiffness of culturing substrates. In agreement with the scientific lit-erature (Barthes et al., 2015; Cichon et al., 2012; Shen et al., 2014;Strehmel et al., 2013; Sunami et al., 2014), our results on the mor-phological properties of A549 cells in 2D/3D cultures indeed confirmedthe influence of three-dimensional spatial organization in defining thecell shape. In detail, shifting from 2D to 3D, cancer cells switched into aspheroidal shape (Fig. 1b). Not surprisingly, A549 cells responded alsoto differences in substrate stiffness by re-organizing their cytoskeleton.Consistent with previous studies (Engler et al., 2006; Shukla et al.,2016), A549 cells cultured on a stiff substrate (glass) showed actinstress fibres formation; whereas, cells grown of softer substrates (Ma-trigel™ and PuraMatrix™) exhibited cortical F-actin. Such cytoskeletonre-organization translated into changes in the mechanical properties ofA549 cells, which showed a smaller Young’s modulus (i.e., stiffness)when cultured in soft matrices (Matrigel™ and PuraMatrix™) (Figs. 5and 6). This result is consistent with previous studies on A549 cells(Shukla et al., 2016) and fibroblasts (Solon et al., 2007). As previouslyelucidated by other researchers (Tee et al., 2011), we suggest that thereduced stiffness detected in cells cultured in 3D soft matrices in ourstudy, was due to the decreased cortical tension associated with the F-actin re-distribution.

Our results on the mechanical properties of A549 cells in 2D/3Dmicroenvironment seemed to highlight that A549 cells cultured in 3Dpossessed higher mesenchymal characteristics. In fact, cell stiffness/Young’s modulus is known to decrease with the gain of metastatic po-tential (Xu et al., 2012) through a power-law relation (Swaminathanet al., 2011). Since decreased substrate stiffness and 3D architectureresulted in a reduction of A549 cells’ Young’s modulus (both in fixedand live specimens) (Figs. 5 and 6), we originally hypothesised that

Fig. 7. Cell viability of untreated cell cultures. Histogram of the ATP levels in untreatedcultures of A549 cells grown for 4 d (from left to right) on glass (G) in 2D, or in 3D inMatrigel™ (M) and PuraMatrix™ (PM). Data are normalized to the ATP levels in 2D cul-tures. Data are reported as average ÷ standard error of the mean (ntests ≥ (5). A pvalue> 0.05 was considered not statistical significant (ns) when compared to the ATPlevels in A549 cells grown on glass (two-way ANOVA with Dunnet post-test).

A. Prina-Mello et al. Tissue and Cell 50 (2018) 15–30

25

these parameters would also alter the expression of EMT markers, suchas E-cadherin, vimentin and fibronectin, towards a mesenchymal phe-notype. However, as shown in Fig. 3, decreasing cell substrate stiffnessor shifting from 2D to 3D cultures did not trigger spontaneous EMT. E-cadherin expression (epithelial marker) in 3D cultures grown in Ma-trigel™ or PuraMatrix™ 0.15% was comparable or even higher to thatfound in 2D cultures, but it could not be detected in PuraMatrix™ 0.25%and 0.30%. The lack of E-cadherin did not correspond, however, to an

enhanced expression of fibronectin (mesenchymal marker). No sig-nificant change in the expression of the mesenchymal marker vimentinwas observed in cells growing on the different substrates. This was inagreement with previous studies on A549 cells (Tilghman et al., 2010).Thus, we concluded that cell substrate stiffness or 3D architecture wasnot correlated to EMT. This finding is consistent with a previous studyby Shuckla et al. showing that substrate stiffness does not regulate EMTin A549 cell line (Shukla et al., 2016). Nevertheless, the enhanced

Fig. 8. Secretion of cytokines, chemokines, EGF and TGFβ-1 and subsequent activation of specific cellular pathways. (a) Secretion levels of 12 extracellular signalling molecules (IL-2, IL-4, IL-5, IL-6, IL-8, IL-10, IL-12 p70, IL-13, IL-17A, TNF-α, GM-CSF and IFN-γ) in A549 cultures grown (from top to bottom) on glass (2D) or in 3D in Matrigel™ or PuraMatrix™ at threedifferent concentrations (0.15%, 0.25% and 0.30%), as detected by PathScan Antibody Array Kit Th1/Th2/Th17. Positive (PT) and negative (NT) controls are included in the array kit.The spot intensity is directly proportional to the amount of target molecule detected. Spot intensity quantification was carried out for evaluating differences in IL-8 secretion levels amongthe different cell cultures. For each culture, IL-8 levels are reported as average ÷ standard deviation (nspotsquantified = 2). (b,c,e,g) Amount of (b) IL-1β, (c) CXCL-1, (e) EGF and (g) TGFβ-1detected by ELISAs in supernatants harvested from A549 cell cultures grown (from left to right) on glass (G), in Matrigel™ (M) or in PuraMatrix™ (PM) at three different concentrations(0.15%, 0.25% and 0.30%). Data are reported as average ÷ standard error of the mean (ntests = 2; nreplicates = 2). (d,f,h) Western blot showing the expression of: (d) phospho-Akt (p-Akt),Akt, phosphor-mTOR (p-mTOR). (f) phospho-Erk1/2 (p-Erk1/2), Erk1/2; and (h) phospho-SMAD2 (p-SMAD2) in A549 cells cultured in (from left to right): 2D on glass substrates (G), or3D in Matrigel™ (M) or PuraMatrix™ (PM) at three different concentrations (0.15%, 0.25% and 0.30%). Two individual experiments (n1 and n2) are presented for each culture. Therelative expression levels of p-Akt, p-mTOR and p-Erk1/2 in each sample was calculated as the ratio between the protein band intensity and that of the loading control (GADPH). (i)Western blot showing the expression of p-SMAD2 in A549 cells grown on glass in the absence (−) and in the presence (+) of recombinant human TGF-β1 (rh TGF-(1) (5 ng/ml). (a–c, e,g) p values>0.05 were considered not statistical significant (ns) when compared to values found for A549 cells grown on glass (2D) (one-way ANOVA followed by Turkey’s multiplecomparison test).

A. Prina-Mello et al. Tissue and Cell 50 (2018) 15–30

26

expression of fibronectin in 3D cultures grown in Matrigel™ and Pur-aMatrix™ 0.15%, which are indeed softer than glass substrates, seemedin conflict with previous studies reporting that soft matrices do notinduce expression of mesenchymal markers (Brown et al., 2013;Markowski et al., 2012; Mason et al., 2012; Tilghman et al., 2010).According to recent studies (Brown et al., 2013; Markowski et al., 2012;Wei and Yang, 2016), EMT is triggered by high substrate stiffnessthrough TGFβ signalling. In our study, the total levels of latent TGFβ(Supplementary Fig. S4) and free-active TGFβ-1 (which actually bindsto the TGFβ receptor and exerts biological functions) (Fig. 8g) werefound to be lower in 3D cell cultures than in 2D. Accordingly, phos-phorylation of SMAD2, which is triggered by TGFβ signalling and in-duces EMT through SNAIL activation, was detected only in 2D cultures(Fig. 8i). Other physical properties of the hydrogels, such as cross-linking and pore size (which can be altered by modifying factors aspolymerization conditions), may therefore have affected the expressionof fibronectin in 3D cell cultures grown in Matrigel™ and PuraMatrix™0.15%. In addition, physical cues are not the only parameters influen-cing the cell phenotype in 3D cultures. According to the manufacturer(Corning), for example, trace amounts of fibronectin are present in theMatrigel™ matrix; thus, the high levels of fibronectin detected by LSCMin 3D cultures grown in such hydrogel could be also partly or com-pletely due to the composition of the culturing microenvironment.Thus, it must be acknowledged that the stiffness of the bulk substratemay not be the only cause of the observed change in EMT biomarkersexpression.

Although no EMT was detected, A549 cells’ membrane was char-acterized by F-actin projections when cultured in 3D (Fig. 1b). Suchcytoskeletal protrusions were more evident in those cultures grown inPuraMatrix™ (Fig. 4b). Topographical data showed that this translatedinto a higher cell surface roughness (as measured by AFM) of A549 cells

grown in 3D in PuraMatrix™ (Fig. 4a). Since PuraMatrix™ is sig-nificantly less stiff than the other substrates tested of at least 75 Pa(Allen et al., 2011), we could conclude that an inverse correlation be-tween the stiffness of the substrate and the surface membrane struc-tures/roughness of the cancer cells grown in 3D was observed. Hence,we hypothesized that, the surface membrane structures found in 3Dcultures grown in the soft matrices (Matrigel™ and PuraMatrix™) wereassociated with the process of migration and invasion. Shulka et al.have in fact shown that matrix stiffness can influence A549 cells mi-gration even when it does not alter the expression of EMT biochemicalmarkers (Shukla et al., 2016). Our results showed that, although ex-pressed in both 2D and 3D cell cultures (Fig. 2b), cortactin was highlyco-localized in correspondence of the F-actin-enriched protrusions onlyin Matrigel™-/PuraMatrix™-based cultures (Fig. 2a). The ECM-de-grading activity of invadopodia is due to the presence of matrix me-talloproteinases (MMPs), and in particular MT1-MMP (Chen and Wang,1999; Poincloux et al., 2009). We found that MT1-MMP was indeedexpressed in 3D cultures grown in Matrigel™ and PuraMatrix™ 0.15%(Fig. 2c). These results, together with those on cells’ stiffness/Young’smodulus, seem indeed to indicate that 3D cultures were characterizedby the formation of invadopodia and therefore by a higher invasionpotential, although this was not linked to EMT phenotypic changes.

Since induction of EMT could not explain the formation of in-vadopodia in 3D cultures, we investigated the biochemical signature ofA549 cells in 2D/3D culturing microenvironments. Our ELISAs results(Fig. 8) were consistent with scientific literature showing that, A549cells are not capable to produce GM-CSF, a mediator that can stimulatethe invasive capacity of human lung cancer cells (Uemura et al., 2006).IL-6 secretion by A549 cells is known to be induced by IL-1β and TNF-α(Crestani et al., 1994), which could not be detected in the cell culturemedia of any of our in vitro model (Fig. 8a). Accordingly, no IL-6

Fig. 9. Cell cultures response to docetaxel. (a) ATP levels and (b) percentage (%) cytotoxicity in A549 cells cultured (from left to right) in 2D on glass, or in 3D in Matrigel™ orPuraMatrix™ (at three different concentrations: 0.15%, 0.25% and 0.30%). Cultures were exposed to 10 increasing concentrations of docetaxel. Data are reported as average ± standarderror of the mean (ATP levels: ntests ≥ 3; cytotoxicity: ntests ≥ (5). The abbreviations NT and PT indicate the negative and positive controls, respectively. The symbols (*), (**) and (***)highlight statistical significance as compared to the corresponding NT (p values < 0.05, 0.01 and 0.001, respectively) (two-way ANOVA followed by Dunnet post-test). (a) Data for eachcell culture are normalized on the corresponding negative control.

A. Prina-Mello et al. Tissue and Cell 50 (2018) 15–30

27

secretion was found in 2D or 3D cultures (Fig. 8a), supporting our datashowing that spontaneous EMT did not occur in response to changes insubstrate/architecture. IL-6 does in fact trigger EMT in A549 cell line(Shintani et al., 2016). Similarly, we could not detect secretion of IL-17p70 (Fig. 8a), which once again induces EMT in NSCLC cell lines (Guet al., 2015; Huang et al., 2016). In addition, in A549 cell line, IFN-γ isreported to efficiently reduce IL-8 secretion under the influence of IL-1β(Boost et al., 2008). In our study, we could not detect any secretion ofIFN-γ (Fig. 8a) or IL-1β (Fig. 8b); accordingly, IL-8 secretion levels wereremarkable, with a spot intensity comparable to that of the assay po-sitive control (PT) (Fig. 8a). Autocrine IL8/CXCL-8, which is known tobe especially elevated in NSCLC cell lines (Liu et al., 2016), facilitatesinvasion, EMT and survival in A549 cells (Desai et al., 2013) and othercancer cells (David et al., 2016; Waugh and Wilson, 2008). The keypathways that are activated in cancer cells in response to IL-8 are three:PI3K, MAPK and Rho-GTPase signalling cascades (Liu et al., 2016;Waugh and Wilson, 2008). The presence of IL-8 supported our con-clusion that cortactin synthesis (and the subsequent formation of in-vadopodia) was activated through Rho-GTPase signalling cascade(Fig. 2b). Notably, IL-8 levels were higher in 3D cell cultures than in 2Dmodels, which made us conclude that IL-8 was responsible for pro-moting the cancer cell invasive potential in 3D cultures independentlyfrom EMT. We proved also that such high IL-8 amounts correlated to apreferential activation of the PI3K/Akt signalling cascade in in vitromodels grown within Matrigel™ and of the MAPK/Erk1/2 pathway inPuraMatrix™-based 3D cultures (Fig. 8d and f, respectively). Both cas-cades promote cancer cells proliferation and survival. This might ex-plain the increased cell viability of untreated 3D cultures grown onPuraMatrix™ 0.30% as compared to 2D models (Fig. 7). Notably, A549cells cultured in 3D in PuraMatrix™ 0.25% and 0.30% showed a lowerchemoresistance to docetaxel, a chemotherapeutic drug used in clinicalsetting for NSCLC treatment, than those cells cultured on glass (2D)(Fig. 9). This is in contradiction with the concept that, on average, theefficacy of anticancer agents is reduced in 3D compared to 2D cultures,as for example shown for A549 cells in Godugu et al. (Godugu et al.,2013). Docetaxel, a Taxol-derived drug, is purposely designed to act oncell cytoskeleton architecture dynamics, preventing microtubules net-work re-organization, inhibiting mitosis and subsequently suppressingmetastasis. Thus, this drug exerts its cytotoxic action by selectivelyaltering the mechanical properties of cancer cells (Suresh, 2007). Thedifference in F-actin organization and mechanical properties between2D and 3D cultures, as well as between 3D models grown in Matrigel™and in PuraMatrix™, could therefore explain the lack of chemoresis-tance detected in our study. However, since resistance to docetaxel ismultifactorial, further studies would be needed to elucidate why thesensitivity of A549 cells did not decrease in 3D cultures grown in softmatrices and whether this is (or is not) a cell type- or substrate-de-pendent effect.

5. Conclusions

In conclusion, in this study we evaluated a range of cellular prop-erties within 3D models (morphology, topography, mechanical andbiochemical signatures) that are key determinants in tumour cells in-vasion and cancer progression and survival in human patients. Thedifferences among cell models described herein, highlight the im-portance of characterizing the properties of 3D cell models used inpreclinical testing, as alternatives to animal studies. As also suggestedby other experts in the field (Berg et al., 2014; Edmondson et al., 2014),we strongly believe that dissecting and quantifying the effect of 3D cellculture substrates on cancer cells phenotype transformations is indeedcritical for drug efficacy assays. 3D cell models that show a satisfactorycorrelation to clinical scenarios could in fact be used for an effectivepreclinical efficacy validation of novel anticancer drugs, ultimatelyproviding a perspective platform for personalised medicine approaches.Our study can therefore find direct application in any attempt to relate

the pathological condition/process of interest with the appropriate 3Dsystem.

Conflict of interest statement

The authors disclose no potential conflicts of interest.

Acknowledgements

The authors thank Prof. Valeria Nicolosi (CRANN/School of Physics,TCD) for technical assistance and kind help, Dr. Despoina Bazou(University College Dublin/Mater Misericordiae Dublin) for usefuldiscussions and suggestions, and the Advanced Microscopy Facility(AML) of TCD. B.L. acknowledges support provided by DGPP (R12242),which is funded under the Programme for Research in Third LevelInstitutions (PRTLI) Cycle 5 and co-funded by the European RegionalDevelopment Fund. J.I.K. and S.P.J acknowledge support from ScienceFoundation Ireland (SFI12/IA/1449). J.I.K. acknowledges supportprovided by the Nanoscale Electrical Metrology project funded by theCollaborative Centre for Applied Nanotechnology (CCAN). DM thanksthe Irish Research Council under the Government of IrelandPostdoctoral Fellowship scheme for funding.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, in theonline version, at https://doi.org/10.1016/j.tice.2017.11.003.

References

Abbott, A., 2003. Cell culture: biology's new dimension. Nature 424, 870–872.Achilli, T.M., Meyer, J., Morgan, J.R., 2012. Advances in the formation, use and under-

standing of multi-cellular spheroids. Expert Opin. Biol. Ther. 12, 1347–1360.Albini, A., Noonan, D.M., 2010. The ‘chemoinvasion’ assay, 25 years and still going

strong: the use of reconstituted basement membranes to study cell invasion and an-giogenesis. Curr. Opin. Cell Biol. 22, 677–689.

Allen, P., Melero-Martin, J., Bischoff, J., 2011. Type I collagen, fibrin and PuraMatrixmatrices provide permissive environments for human endothelial and mesenchymalprogenitor cells to form neovascular networks. J. Tissue Eng. Regen. Med. 5, e74–86.

Aparicio, S., Hidalgo, M., Kung, A.L., 2015. Examining the utility of patient-derived xe-nograft mouse models. Nat. Rev. Cancer 15, 311–316.

Arrowsmith, J., Miller, P., 2013. Trial watch: phase II and phase III attrition rates2011–2012. Nat. Rev. Drug Discov. 12, 569.

Arrowsmith, J., 2011. Trial watch: phase II failures: 2008–2010. Nat. Rev. Drug Discov.10, 328–329.

Barthes, J., Vrana, N.E., Ozcelik, H., Gahoual, R., Francois, Y.N., Bacharouche, J.,Francius, G., Hemmerle, J., Metz-Boutigue, M.H., Schaaf, P., Lavalle, P., 2015.Priming cells for their final destination: microenvironment controlled cell culture bya modular ECM-mimicking feeder film. Biomater. Sci. 3, 1302–1311.

Bazou, D., Behan, G., Reid, C., Boland, J.J., Zhang, H.Z., 2011. Imaging of human coloncancer cells using He-ion scanning microscopy. J. Microsc. 242, 290–294.

Bell, D.C., 2009. Contrast mechanisms and image formation in helium ion microscopy.Microsc. Microanal. 15, 147–153.

Berg, L.E., Hsu, Y.C., Lee, J.A., 2014. Consideration of the cellular microenvironment:physiologically relevant co-culture systems in drug discovery. Adv. Drug Deliv. Rev.69–70, 190–204.

Boost, K.A., Sadik, C.D., Bachmann, M., Zwissler, B., Pfeilschifter, J., Muhl, H., 2008. IFN-gamma impairs release of IL-8 by IL-1beta-stimulated A549 lung carcinoma cells.BMC Cancer 8, 265.

Braet, F., De Zanger, R., Wisse, E., 1997. Drying cells for SEM, AFM and TEM by hex-amethyldisilazane: a study on hepatic endothelial cells. J. Microsc. 186, 84–87.

Brown, A.C., Fiore, V.F., Sulchek, T.A., Barker, T.H., 2013. Physical and chemical mi-croenvironmental cues orthogonally control the degree and duration of fibrosis-as-sociated epithelial-to-mesenchymal transitions. J. Pathol. 229, 25–35.

Buxboim, A., Rajagopal, K., Brown, A.E., Discher, D.E., 2010. How deeply cells feel:methods for thin gels. J. Phys. Condens. Matter 22, 194116.

Caino, M.C., Lopez-Haber, C., Kissil, J.L., Kazanietz, M.G., 2012. Non-small cell lungcarcinoma cell motility, rac activation and metastatic dissemination are mediated byprotein kinase C epsilon. PLoS One 7, e31714.

Chen, W.T., Wang, J.Y., 1999. Specialized surface protrusions of invasive cells, in-vadopodia and lamellipodia, have differential MT1-MMP, MMP-2, and TIMP-2 lo-calization. Ann. N. Y. Acad. Sci. 878, 361–371.

Cichon, M.A., Gainullin, V.G., Zhang, Y., Radisky, D.C., 2012. Growth of lung cancer cellsin three-dimensional microenvironments reveals key features of tumor malignancy.Integr. Biol. (Camb.) 4, 440–448.

Clark, E.S., Whigham, A.S., Yarbrough, W.G., Weaver, A.M., 2007. Cortactin is an

A. Prina-Mello et al. Tissue and Cell 50 (2018) 15–30

28

essential regulator of matrix metalloproteinase secretion and extracellular matrixdegradation in invadopodia. Cancer Res. 67, 4227–4235.

Cook, D., Brown, D., Alexander, R., March, R., Morgan, P., Satterthwaite, G., Pangalos,M.N., 2014. Lessons learned from the fate of AstraZeneca's drug pipeline: a five-dimensional framework. Nat. Rev. Drug Discov. 13, 419–431.

Corning, https://www.corning.com/media/worldwide/cls/documents/CLS-DL-CC-026%20DL.pdf (Last access: Aug. 2017).

Crestani, B., Cornillet, P., Dehoux, M., Rolland, C., Guenounou, M., Aubier, M., 1994.Alveolar type II epithelial cells produce interleukin-6 in vitro and in vivo: regulationby alveolar macrophage secretory products. J. Clin. Invest. 94, 731–740.

Cross, S.E., Jin, Y.S., Rao, J., Gimzewski, J.K., 2007. Nanomechanical analysis of cellsfrom cancer patients. Nat. Nanotechnol. 2, 780–783.

Cross, S.E., Jin, Y.S., Tondre, J., Wong, R., Rao, J., Gimzewski, J.K., 2008. AFM-basedanalysis of human metastatic cancer cells. Nanotechnology 19, 384003.

David, J.M., Dominguez, C., Hamilton, D.H., Palena, C., 2016. The IL-8/IL-8R axis: adouble agent in tumor immune resistance. Vaccines (Basel) 4, 22.

Desai, S., Laskar, S., Pandey, B.N., 2013. Autocrine IL-8 and VEGF mediate epithelial-mesenchymal transition and invasiveness via p38/JNK-ATF-2 signalling in A549 lungcancer cells. Cell. Signal. 25, 1780–1791.

Discher, D.E., Janmey, P., Wang, Y.L., 2005. Tissue cells feel and respond to the stiffnessof their substrate. Science 310, 1139–1143.

Dokukin, M.E., Guz, N.V., Gaikwad, R.M., Woodworth, C.D., Sokolov, I., 2011. Cell sur-face as a fractal: normal and cancerous cervical cells demonstrate different fractalbehavior of surface adhesion maps at the nanoscale. Phys. Rev. Lett. 107, 028101.

Dufrene, Y.F., Martinez-Martin, D., Medalsy, I., Alsteens, D., Muller, D.J., 2013.Multiparametric imaging of biological systems by force-distance curve-based AFM.Nat. Methods 10, 847–854.

Edmondson, R., Broglie, J.J., Adcock, A.F., Yang, L., 2014. Three-dimensional cell culturesystems and their applications in drug discovery and cell-based biosensors. AssayDrug Dev. Technol. 12, 207–218.

Egeblad, M., Nakasone, E.S., Werb, Z., 2010. Tumors as organs: complex tissues thatinterface with the entire organism. Dev. Cell 18, 884–901.

Engler, A.J., Sen, S., Sweeney, H.L., Discher, D.E., 2006. Matrix elasticity directs stem celllineage specification. Cell 126, 677–689.

Ferlay, J., Shin, H.R., Bray, F., Forman, D., Mathers, C., Parkin, D.M., 2010. Estimates ofworldwide burden of cancer in 2008: GLOBOCAN 2008. Int. J. Cancer 127,2893–2917.

Fitzgerald, K.A., Malhotra, M., Curtin, C.M., O' Brien, F.J., O' Driscoll, C.M., 2015. Life in3D is never flat: 3D models to optimise drug delivery. J. Control. Release 215, 39–54.

Friedl, P., Sahai, E., Weiss, S., Yamada, K.M., 2012. New dimensions in cell migration.Nat. Rev. Mol. Cell Biol. 13, 743–747.

Friedrich, J., Ebner, R., Kunz-Schughart, L.A., 2007. Experimental anti-tumor therapy in3-D: spheroids–old hat or new challenge? Int. J. Radiat. Biol. 83, 849–871.

Friedrich, J., Seidel, C., Ebner, R., Kunz-Schughart, L.A., 2009. Spheroid-based drugscreen: considerations and practical approach. Nat. Protoc. 4, 309–324.

Gazdar, A.F., Girard, L., Lockwood, W.W., Lam, W.L., Minna, J.D., 2010. Lung cancer celllines as tools for biomedical discovery and research. J. Natl. Cancer Inst. 102,1310–1321.

Georges, P.C., Janmey, P.A., 2005. Cell type-specific response to growth on soft materials.J. Appl. Physiol. (1985) 98, 1547–1553.

Gillet, J.P., Varma, S., Gottesman, M.M., 2013. The clinical relevance of cancer cell lines.J. Natl. Cancer Inst. 105, 452–458.

Godugu, C., Patel, A.R., Desai, U., Andey, T., Sams, A., Singh, M., 2013. AlgiMatrix based3D cell culture system as an in-vitro tumor model for anticancer studies. PLoS One 8,e53708.

Gu, K., Li, M.M., Shen, J., Liu, F., Cao, J.Y., Jin, S., Yu, Y., 2015. Interleukin-17-inducedEMT promotes lung cancer cell migration and invasion via NF-kappaB/ZEB1 signalpathway. Am. J. Cancer. Res. 5, 1169–1179.

Guz, N.V., Dokukin, M.E., Woodworth, C.D., Cardin, A., Sokolov, I., 2015. Towards earlydetection of cervical cancer: fractal dimension of AFM images of human cervicalepithelial cells at different stages of progression to cancer. Nanomedicine 11,1667–1675.

Hackam, D.G., 2007. Translating animal research into clinical benefit. BMJ 334, 163–164.Hakanson, M., Cukierman, E., Charnley, M., 2014. Miniaturized pre-clinical cancer

models as research and diagnostic tools. Adv. Drug Deliv. Rev. 69–70, 52–66.Hartung, T., 2008. Food for thought… on animal tests. ALTEX 25, 3–16.Hartung, T., 2017. Perspectives on in vitro to in vivo extrapolations. Appl. In Vitro Toxicol.

http://dx.doi.org/10.1089/aivt.2016.0026. Ahead of print.Hay, M., Thomas, D.W., Craighead, J.L., Economides, C., Rosenthal, J., 2014. Clinical

development success rates for investigational drugs. Nat. Biotechnol. 32, 40–51.Hirschhaeuser, F., Menne, H., Dittfeld, C., West, J., Mueller-Klieser, W., Kunz-Schughart,

L.A., 2010. Multicellular tumor spheroids: an underestimated tool is catching upagain. J. Biotechnol. 148, 3–15.

Horvath, P., Aulner, N., Bickle, M., Davies, A.M., Nery, E.D., Ebner, D., Montoya, M.C.,Ostling, P., Pietiainen, V., Price, L.S., Shorte, S.L., Turcatti, G., von Schantz, C.,Carragher, N.O., 2016. Screening out irrelevant cell-based models of disease. Nat.Rev. Drug Discov. 15, 751–769.

Huang, Q., Han, J., Fan, J., Duan, L., Guo, M., Lv, Z., Hu, G., Chen, L., Wu, F., Tao, X., Xu,J., Jin, Y., 2016. IL-17 induces EMT via Stat3 in lung adenocarcinoma. Am. J. Cancer.Res. 6, 440–451.

Hutmacher, D.W., 2010. Biomaterials offer cancer research the third dimension. Nat.Mater. 9, 90–93.

Iyer, S., Gaikwad, R.M., Subba-Rao, V., Woodworth, C.D., Sokolov, I., 2009. Atomic forcemicroscopy detects differences in the surface brush of normal and cancerous cells.Nat. Nanotechnol. 4, 389–393.

Jaafar, H., Sharif, S.E., Murtey, M.D., 2012. Distinctive features of advancing breast

cancer cells and interactions with surrounding stroma observed under the scanningelectron microscope. Asian Pac. J. Cancer Prev. 13, 1305–1310.

Joens, M.S., Huynh, C., Kasuboski, J.M., Ferranti, D., Sigal, Y.J., Zeitvogel, F., Obst, M.,Burkhardt, C.J., Curran, K.P., Chalasani, S.H., Stern, L.A., Goetze, B., Fitzpatrick, J.A.,2013. Helium Ion Microscopy (HIM) for the imaging of biological samples at sub-nanometer resolution. Sci. Rep. 3, 3514.

Kamb, A., 2005. What's wrong with our cancer models? Nat. Rev. Drug Discov. 4,161–165.

Kankaanpaa, P., Paavolainen, L., Tiitta, S., Karjalainen, M., Paivarinne, J., Nieminen, J.,Marjomaki, V., Heino, J., White, D.J., 2012. BioImageXD: an open, general-purposeand high-throughput image-processing platform. Nat. Methods 9, 683–689.

Kimlin, L.C., Casagrande, G., Virador, V.M., 2013. In vitro three-dimensional (3D) modelsin cancer research: an update. Mol. Carcinog. 52, 167–182.

Kola, I., Landis, J., 2004. Can the pharmaceutical industry reduce attrition rates? Nat.Rev. Drug Discov. 3, 711–715.

Kural, M.H., Billiar, K.L., 2013. Regulating tension in three-dimensional culture en-vironments. Exp. Cell Res. 319, 2447–2459.

Landry, J.J., Pyl, P.T., Rausch, T., Zichner, T., Tekkedil, M.M., Stutz, A.M., Jauch, A.,Aiyar, R.S., Pau, G., Delhomme, N., Gagneur, J., Korbel, J.O., Huber, W., Steinmetz,L.M., 2013. The genomic and transcriptomic landscape of a HeLa cell line. G3(Bethesda) 3, 1213–1224.

Lee, J.T., Chow, K.L., 2012. SEM sample preparation for cells on 3D scaffolds by freeze-drying and HMDS. Scanning 34, 12–25.

Lekka, M., Laidler, P., 2009. Applicability of AFM in cancer detection. Nat. Nanotechnol.4 (72) (author reply 72–73).

Lieberman, M.M., Patterson, G.M., Moore, R.E., 2001. In vitro bioassays for anticancerdrug screening: effects of cell concentration and other assay parameters on growthinhibitory activity. Cancer Lett. 173, 21–29.

Liu, H., Tan, Q., Geddie, W.R., Jewett, M.A., Phillips, N., Ke, D., Simmons, C.A., Sun, Y.,2014. Biophysical characterization of bladder cancer cells with different metastaticpotential. Cell Biochem. Biophys. 68, 241–246.

Liu, Q., Li, A., Tian, Y., Wu, J.D., Liu, Y., Li, T., Chen, Y., Han, X., Wu, K., 2016. TheCXCL8-CXCR1/2 pathways in cancer. Cytokine Growth Factor Rev. 31, 61–71.

Lo, H.M., Lai, T.H., Li, C.H., Wu, W.B., 2014. TNF-alpha induces CXCL1 chemokine ex-pression and release in human vascular endothelial cells in vitro via two distinctsignaling pathways. Acta Pharmacol. Sin. 35, 339–350.

Mak, I.W., Evaniew, N., Ghert, M., 2014. Lost in translation: animal models and clinicaltrials in cancer treatment. Am. J. Transl. Res. 6, 114–118.

Malvezzi, M., Bertuccio, P., Rosso, T., Rota, M., Levi, F., La Vecchia, C., Negri, E., 2015.European cancer mortality predictions for the year 2015: does lung cancer have thehighest death rate in EU women? Ann. Oncol. 26, 779–786.

Markowski, M.C., Brown, A.C., Barker, T.H., 2012. Directing epithelial to mesenchymaltransition through engineered microenvironments displaying orthogonal adhesiveand mechanical cues. J. Biomed. Mater. Res. A 100, 2119–2127.

Mason, B.N., Califano, J.P., Reinhart-King, C.A., 2012. Matrix stiffness: a regulator ofcellular behavior and tissue formation. In: Bhatia, S.K. (Ed.), EngineeringBiomaterials for Regenerative Medicine. Springer, New York, pp. 19–37.

Mehta, G., Hsiao, A.Y., Ingram, M., Luker, G.D., Takayama, S., 2012. Opportunities andchallenges for use of tumor spheroids as models to test drug delivery and efficacy. J.Control. Release 164, 192–204.

Morton, K.C., Baker, L.A., 2014. Atomic force microscopy-based bioanalysis for the studyof disease. Anal. Methods 6, 4932–4955.

Nyga, A., Cheema, U., Loizidou, M., 2011. 3D tumour models: novel in vitro approachesto cancer studies. J. Cell Commun. Signal. 5, 239–248.

Olson, H., Betton, G., Robinson, D., Thomas, K., Monro, A., Kolaja, G., Lilly, P., Sanders,J., Sipes, G., Bracken, W., Dorato, M., Van Deun, K., Smith, P., Berger, B., Heller, A.,2000. Concordance of the toxicity of pharmaceuticals in humans and in animals.Regul. Toxicol. Pharmacol. 32, 56–67.

Pickup, M.W., Mouw, J.K., Weaver, V.M., 2014. The extracellular matrix modulates thehallmarks of cancer. EMBO Rep. 15, 1243–1253.

Poincloux, R., Lizarraga, F., Chavrier, P., 2009. Matrix invasion by tumour cells: a focuson MT1-MMP trafficking to invadopodia. J. Cell Sci. 122, 3015–3024.

Poondru, S., Parchment, R.E., Purohit, V., LoRusso, P., Horwitz, J.P., Hazeldine, S.T.,Polin, L., Corbett, T., Jasti, B.R., 2002. Lack of in vitro-in vivo correlation of a novelinvestigational anticancer agent, SH 30. Invest. New Drugs 20, 23–33.

Pound, P., Ebrahim, S., Sandercock, P., Bracken, M.B., Roberts, I., Reviewing AnimalTrials G. Systematically, 2004. Where is the evidence that animal research benefitshumans? BMJ 328, 514–517.

Radmacher, M., 1997. Measuring the elastic properties of biological samples with theAFM. IEEE Eng. Med. Biol. Mag. 16, 47–57.

Rangarajan, A., Weinberg, R.A., 2003. Opinion: comparative biology of mouse versushuman cells: modelling human cancer in mice. Nat. Rev. Cancer 3, 952–959.

Rehfeldt, F., Engler, A.J., Eckhardt, A., Ahmed, F., Discher, D.E., 2007. Cell responses tothe mechanochemical microenvironment–implications for regenerative medicine anddrug delivery. Adv. Drug Deliv. Rev. 59, 1329–1339.

Rice, W.L., Van Hoek, A.N., Paunescu, T.G., Huynh, C., Goetze, B., Singh, B., Scipioni, L.,Stern, L.A., Brown, D., 2013. High resolution helium ion scanning microscopy of therat kidney. PLoS One 8, e57051.

Rodriguez-Enriquez, S., Gallardo-Perez, J.C., Aviles-Salas, A., Marin-Hernandez, A.,Carreno-Fuentes, L., Maldonado-Lagunas, V., Moreno-Sanchez, R., 2008. Energymetabolism transition in multi-cellular human tumor spheroids. J. Cell. Physiol. 216,189–197.

Ruggeri, B.A., Camp, F., Miknyoczki, S., 2014. Animal models of disease: pre-clinicalanimal models of cancer and their applications and utility in drug discovery.Biochem. Pharmacol. 87, 150–161.

Sader, J.E., Chon, J.W.M., Mulvaney, P., 1999. Calibration of rectangular atomic force

A. Prina-Mello et al. Tissue and Cell 50 (2018) 15–30

29

microscope cantilevers. Rev. Sci. Instrum. 70, 3967–3969.Shen, Y.I., Abaci, H.E., Krupsi, Y., Weng, L.C., Burdick, J.A., Gerecht, S., 2014. Hyaluronic

acid hydrogel stiffness and oxygen tension affect cancer cell fate and endothelialsprouting. Biomater. Sci. 2, 655–665.

Shieh, J.M., Tsai, Y.J., Tsou, C.J., Wu, W.B., 2014. CXCL1 regulation in human pulmonaryepithelial cells by tumor necrosis factor. Cell. Physiol. Biochem. 34, 1373–1384.

Shield, K., Ackland, M.L., Ahmed, N., Rice, G.E., 2009. Multicellular spheroids in ovariancancer metastases: biology and pathology. Gynecol. Oncol. 113, 143–148.

Shintani, Y., Fujiwara, A., Kimura, T., Kawamura, T., Funaki, S., Minami, M., Okumura,M., 2016. IL-6 secreted from cancer-associated fibroblasts mediates chemoresistancein NSCLC by increasing epithelial-mesenchymal transition signaling. J. Thorac.Oncol. 11, 1482–1492.

Shukla, V.C., Higuita-Castro, N., Nana-Sinkam, P., Ghadiali, S.N., 2016. Substrate stiffnessmodulates lung cancer cell migration but not epithelial to mesenchymal transition. J.Biomed. Mater. Res. A 104, 1182–1193.

Smalley, K.S., Lioni, M., Herlyn, M., 2006. Life isn't flat: taking cancer biology to the nextdimension. In Vitro Cell. Dev. Biol. Anim. 42, 242–247.

Sneddon, I.N., 1965. The relation between load and penetration in the axisymmetricboussinesq problem for a punch of arbitrary profile. Int. J. Eng. Sci. 3, 47–57.

Sokolov, I., 2007. Atomic force microscopy in cancer cell research. In: Singh, N.H.,Thomas, W. (Eds.), Cancer Nanotechnology, pp. 1–17.

Solon, J., Levental, I., Sengupta, K., Georges, P.C., Janmey, P.A., 2007. Fibroblast adap-tation and stiffness matching to soft elastic substrates. Biophys. J. 93, 4453–4461.

Stratmann, A.T., Fecher, D., Wangorsch, G., Gottlich, C., Walles, T., Walles, H., Dandekar,T., Dandekar, G., Nietzer, S.L., 2014. Establishment of a human 3D lung cancer modelbased on a biological tissue matrix combined with a Boolean in silico model. Mol.Oncol. 8, 351–365.

Strehmel, C., Zhang, Z., Strehmel, N., Lensen, M.C., 2013. Cell phenotypic changes ofmouse connective tissue fibroblasts (L-929) to poly(ethylene glycol)-based gels.Biomater. Sci. 1, 850–859.

Sunami, H., Yokota, I., Igarashi, Y., 2014. Influence of the pattern size of micropatternedscaffolds on cell morphology, proliferation, migration and F-actin expression.Biomater. Sci. 2, 399–409.

Suresh, S., 2007. Biomechanics and biophysics of cancer cells. Acta Biomater. 3, 413–438.Swaminathan, V., Mythreye, K., O'Brien, E.T., Berchuck, A., Blobe, G.C., Superfine, R.,

2011. Mechanical stiffness grades metastatic potential in patient tumor cells and incancer cell lines. Cancer Res. 71, 5075–5080.

Tee, S.Y., Fu, J., Chen, C.S., Janmey, P.A., 2011. Cell shape and substrate rigidity bothregulate cell stiffness. Biophys. J. 100, L25–27.

Thiery, J.P., 2002. Epithelial-mesenchymal transitions in tumour progression. Nat. Rev.Cancer 2, 442–454.

Tilghman, R.W., Cowan, C.R., Mih, J.D., Koryakina, Y., Gioeli, D., Slack-Davis, J.K.,Blackman, B.R., Tschumperlin, D.J., Parsons, J.T., 2010. Matrix rigidity regulatescancer cell growth and cellular phenotype. PLoS One 5, e12905.

Touhami, A., Nysten, B., Dufrêne, Y.F., 2003. Nanoscale mapping of the elasticity ofmicrobial cells by atomic force microscopy. Langmuir 19, 4539–4543.

Tsvetkova, E., Goss, G.D., 2012. Drug resistance and its significance for treatment deci-sions in non-small-cell lung cancer. Curr. Oncol. 19, S45–51.

Uemura, Y., Kobayashi, M., Nakata, H., Kubota, T., Bandobashi, K., Saito, T., Taguchi, H.,

2006. Effects of GM-CSF and M-CSF on tumor progression of lung cancer: roles ofMEK1/ERK and AKT/PKB pathways. Int. J. Mol. Med. 18, 365–373.

Unger, C., Kramer, N., Walzl, A., Scherzer, M., Hengstschlager, M., Dolznig, H., 2014.Modeling human carcinomas: physiologically relevant 3D models to improve anti-cancer drug development. Adv. Drug Deliv. Rev. 79–80, 50–67.

van der Worp, H.B., Howells, D.W., Sena, E.S., Porritt, M.J., Rewell, S., O'Collins, V.,Macleod, M.R., 2010. Can animal models of disease reliably inform human studies?PLoS Med. 7, e1000245.

Vinci, M., Gowan, S., Boxall, F., Patterson, L., Zimmermann, M., Court, W., Lomas, C.,Mendiola, M., Hardisson, D., Eccles, S.A., 2012. Advances in establishment andanalysis of three-dimensional tumor spheroid-based functional assays for target va-lidation and drug evaluation. BMC Biol. 10, 29.

Watanabe, T., Kuramochi, H., Takahashi, A., Imai, K., Katsuta, N., Nakayama, T., Fujiki,H., Suganuma, M., 2012. Higher cell stiffness indicating lower metastatic potential inB16 melanoma cell variants and in (−)-epigallocatechin gallate-treated cells. J.Cancer Res. Clin. Oncol. 138, 859–866.

Waugh, D.J., Wilson, C., 2008. The interleukin-8 pathway in cancer. Clin. Cancer Res. 14,6735–6741.

Weaver, V.M., Petersen, O.W., Wang, F., Larabell, C.A., Briand, P., Damsky, C., Bissell,M.J., 1997. Reversion of the malignant phenotype of human breast cells in three-dimensional culture and in vivo by integrin blocking antibodies. J. Cell Biol. 137,231–245.

Wei, S.C., Yang, J., 2016. Forcing through tumor metastasis: the interplay between tissuerigidity and epithelial-mesenchymal transition. Trends Cell Biol. 26, 111–120.

Weiswald, L.B., Richon, S., Massonnet, G., Guinebretiere, J.M., Vacher, S., Laurendeau, I.,Cottu, P., Marangoni, E., Nemati, F., Validire, P., Bellet, D., Bieche, I., Dangles-Marie,V., 2013. A short-term colorectal cancer sphere culture as a relevant tool for humancancer biology investigation. Br. J. Cancer 108, 1720–1731.

Worthington, P., Pochan, D.J., Langhans, S.A., 2015. Peptide hydrogels – versatile ma-trices for 3D cell culture in cancer medicine. Front. Oncol. 5, 92.

Xu, W., Mezencev, R., Kim, B., Wang, L., McDonald, J., Sulchek, T., 2012. Cell stiffness isa biomarker of the metastatic potential of ovarian cancer cells. PLoS One 7, e46609.

Yamada, K.M., Cukierman, E., 2007. Modeling tissue morphogenesis and cancer in 3D.Cell 130, 601–610.

Yang, W., Soares, J., Greninger, P., Edelman, E.J., Lightfoot, H., Forbes, S., Bindal, N.,Beare, D., Smith, J.A., Thompson, I.R., Ramaswamy, S., Futreal, P.A., Haber, D.A.,Stratton, M.R., Benes, C., McDermott, U., Garnett, M.J., 2013. Genomics of DrugSensitivity in Cancer (GDSC): a resource for therapeutic biomarker discovery incancer cells. Nucleic Acids Res. 41, D955–961.

Yilmaz, M., Christofori, G., 2009. EMT, the cytoskeleton, and cancer cell invasion. CancerMetastasis Rev. 28, 15–33.

Zhong, L., Roybal, J., Chaerkady, R., Zhang, W., Choi, K., Alvarez, C.A., Tran, H.,Creighton, C.J., Yan, S., Strieter, R.M., Pandey, A., Kurie, J.M., 2008. Identification ofsecreted proteins that mediate cell-cell interactions in an in vitro model of the lungcancer microenvironment. Cancer Res. 68, 7237–7245.

Zhou, Z., Zheng, C., Li, S., Zhou, X., Liu, Z., He, Q., Zhang, N., Ngan, A., Tang, B., Wang,A., 2013. AFM nanoindentation detection of the elastic modulus of tongue squamouscarcinoma cells with different metastatic potentials. Nanomedicine 9, 864–874.

A. Prina-Mello et al. Tissue and Cell 50 (2018) 15–30

30