dual targeting of mesenchymal and amoeboid...

Cell Death and Survival

Dual Targeting of Mesenchymal and AmoeboidMotility Hinders Metastatic BehaviorBrandon C. Jones1, Laura C. Kelley2, Yuriy V. Loskutov2, Kristina M. Marinak2,Varvara K. Kozyreva2, Matthew B. Smolkin3, and Elena N. Pugacheva1,2

Abstract

Commonly upregulated in human cancers, the scaffoldingprotein NEDD9/HEF1 is a known regulator of mesenchymalmigration and cancer cell plasticity. However, the functionalrole of NEDD9 as a regulator of different migration/invasionmodes in the context of breast cancer metastasis is currentlyunknown. Here, it is reported that NEDD9 is necessary for bothmesenchymal and amoeboid individual cell migration/inva-sion in triple-negative breast cancer (TNBC). NEDD9 deficiencyresults in acquisition of the amoeboid morphology, but severe-ly limits all types of cell motility. Mechanistically, NEDD9promotes mesenchymal migration via VAV2-dependent Rac1activation, and depletion of VAV2 impairs the ability ofNEDD9 to activate Rac1. In addition, NEDD9 supports amesenchymal phenotype through stimulating polymerizationof actin via promoting CTTN phosphorylation in an AURKA-dependent manner. Interestingly, an increase in RhoA activityin NEDD9-depleted cells does not facilitate a switch to func-

tional amoeboid motility, indicating a role of NEDD9 in theregulation of downstream RhoA signaling effectors. Simulta-neous depletion of NEDD9 or inhibition of AURKA in com-bination with inhibition of the amoeboid driver ROCK resultsin an additional decrease in cancer cell migration/invasion.Finally, we confirmed that a dual targeting strategy is a viableand efficient therapeutic approach to hinder the metastasis ofbreast cancer in xenograft models, showcasing the importantneed for further clinical evaluation of this regimen to impedethe spread of disease and improve patient survival.

Implications: This study provides new insight into the thera-peutic benefit of combining NEDD9 depletion with ROCKinhibition to reduce tumor cell dissemination and discovers anew regulatory role of NEDD9 in the modulation of VAV2-dependent activation of Rac1 and actin polymerization. MolCancer Res; 15(6); 670–82. �2017 AACR.

IntroductionScaffolding protein NEDD9/HEF1, commonly overexpressed

in many human cancers, is a critical regulator of cancer cellmigration (1, 2). Cancer cells are capable of migrating eithercollectively or individually (3–5). In collective migration, cellsretain most of their epithelial characteristics including cell–celladhesions and invade as a sheet (6), thus enabling regionalinvasion and lymphatic-based metastasis. Alternatively, individ-ual migration is distinguished by a loss of cell–cell adhesions andthe ability to invade via two different methods known as mes-enchymal and amoeboid (7–9), leading to blood-borne metas-tasis. Depending on the extracellular matrix (ECM) compositionand rigidity of the tumor microenvironment, the mesenchymal

and amoeboid modes can be used interchangeably by the samecell population (3, 8). Initial nascent adhesions are highly presentin amoeboid cells, consisting of integrin, talin, and nonpho-sphorylated paxillin. The recruitment of vinculin and FAK/Srccomplex and activation/phosphorylation of paxillin and FAK isindicative of focal adhesion (FA) maturation, which can culmi-nate in formation of very elongated fibrillar adhesions (10),which are often found in mesenchymal cells.

Elevated expression of NEDD9 protein is required for individ-ual mesenchymal migration of diverse tumor cells (1, 11, 12).Mesenchymal migration is characterized by an elongated cellmorphology, multiple focal/3D adhesions, and the ability todegrade ECM by matrix metalloproteinases (MMP; refs. 7, 13),creating a path through the basement membrane/tissue. Amasterregulator ofmesenchymalmigration, Rac1GTPase, is activated bya number of guanine nucleotide exchange factors (GEF), includ-ing melanoma-specific DOCK3 (14), which is recruited/activatedby NEDD9 (11). Rac1 activates the WAVE-regulatory complexleading to polymerization of actin filaments (15). We havepreviously shown thatNEDD9also activates and stabilizes AuroraA kinase (AURKA; ref. 16), which phosphorylates and activateshistone deacetylase HDAC6 (17), leading to deacetylation andsubsequent increased activity of the actin branching/stabilizingprotein cortactin (CTTN; ref. 18), which is necessary for lamelli-podia and invadopodia formation, structures that define mesen-chymal invasion.

Rac1-driven mesenchymal migration runs counter to the cor-responding RhoA GTPase and ROCKII kinase–driven amoeboid

1Department of Biochemistry, West Virginia University School of Medicine,Morgantown, West Virginia. 2West Virginia University Cancer Institute, WestVirginia University School ofMedicine,Morgantown,WestVirginia. 3Departmentof Pathology, West Virginia University School of Medicine, Morgantown, WestVirginia.

Note: Supplementary data for this article are available at Molecular CancerResearch Online (http://mcr.aacrjournals.org/).

Corresponding Author: Elena N. Pugacheva, Department of Biochemistry andWest Virginia University Cancer Institute, 1 Medical Center Drive, West VirginiaUniversity School of Medicine, Morgantown, WV 26506. Phone: 304-293-5295;Fax: 304-293-4667; E-mail: [email protected]

doi: 10.1158/1541-7786.MCR-16-0411

�2017 American Association for Cancer Research.

MolecularCancerResearch

Mol Cancer Res; 15(6) June 2017670

on June 26, 2018. © 2017 American Association for Cancer Research. mcr.aacrjournals.org Downloaded from

Published OnlineFirst February 24, 2017; DOI: 10.1158/1541-7786.MCR-16-0411

http://crossmark.crossref.org/dialog/?doi=10.1158/1541-7786.MCR-16-0411&domain=pdf&date_stamp=2017-5-11

http://mcr.aacrjournals.org/

pathway. Amoeboid migration is characterized by a round-cellmorphology, small/weak adhesions, and increased cell con-tractility (7, 19). This type of locomotion does not typically relyon matrix degradation, and instead slightly deforms the cellshape to squeeze through preexisting gaps in the ECM (20).Because of this, amoeboid movement is more suited for inva-sion through less dense and more elastic matrices such ascollagens. Mechanistically, RhoA binds and activates ROCKIIwhich phosphorylates myosin II light chain (MLC2), thusincreasing actomyosin contractility and amoeboid invasion.NEDD9 contributes to downregulation of the amoeboid path-way through binding/activating Src kinase, leading to phos-phorylation/inhibition of ROCKII (11).

The impact ofNEDD9on collective tumor cell invasion has notbeen explored. However, collective migration of lymphocytes inNEDD9 knockout mice was shown to be significantly reduced(21), and neuronal tube formation in mouse and chick embryoswas inhibiteduponNEDD9depletion (22).Wehave shown that adecrease in NEDD9 expression leads to significant deficiency inactivation of MMPs (23, 24), suggesting a deficiency of these cellsto create tunnels through the matrix.

As individual tumor cells are capable of utilizing either one ofthese modes of movement, simultaneous targeting of both path-ways may provide significant clinical benefit against metastaticcancers (14, 25). In this current study, we define the molecularmechanisms of NEDD9-dependent Rac1 activation and the effi-cacy of combined inhibition of both types of cell movement tohinder the metastasis of triple-negative breast cancer (TNBC).

Materials and MethodsCell lines, RNAs, and plasmids

TNBC cell lines MDA-MB-231LN (Caliper); Hs578T,HCC1143, BT549, HCC1395 (ATCC); and SUM159 (gift fromDr. Boerner, Karmanos Cancer Institute, Rochester Hills,MI)wereauthenticated and propagated for less than 6 months. si/sh/sgRNAs and plasmids used in this study are described in detailin the Supplementary Materials and Methods. The plasmids/RNAs were introduced via nucleofection (Amaxa) according tomanufacturer's recommendation or virus infection (293T pack-aging cells, Life Technologies) as previously described (23).

Recombinant protein productionRecombinant AURKA (18), wild-type and N/C-term CTTN

(26), and GST and GST-PBD proteins were produced as describedpreviously (27).

Cell morphology analysisCells (104) were plated on two-dimensional (2D) or in three-

dimensional (3D) 1.5mg/mL collagen (AdvancedBioMatrix) or 4mg/mL Matrigel (Corning) as described previously (28). Cellswere treated with 50 or 100 mmol/L Y-27632 (ApexBio) whereindicated. Cells were bright-field imaged on a Leica DM/ILmicro-scope. Cell elongation was measured as a function of cell lengthdivided by cell width (ImageJ software/NIH).

Western blottingWestern blots were performed as described previously (23).

Further details and a list of antibodies used can be found in theSupplementary Materials and Methods.

ImmunofluorescenceThe edges of 18-mmround glass coverslips (Fisher)were coated

with a Super-HT-PAP-PEN hydrophobic slide marker and coatedinside with 0.01% poly-L-lysine (Electron Microscopy Sciences).For 3Dmatrix embedding, 2.5� 104 cells suspended in 200 mL of1.5 mg/mL collagen-I or 4 mg/mL Matrigel plug were loaded onthe coverslips. Coverslips were placed into a 12-well plate andincubated in a 37�C humid chamber during polymerization. Fullmedia were added, and the cells were allowed to acclimate to thematrix for 48 hours. Fixation and immunofluorescence stainingwas carried out as described previously (29). The list of usedantibodies and quantification details are provided in the Supple-mentary Materials and Methods.

Collagen gel contraction assayA total of 1�106 cells per samplewere suspended in1.5mg/mL

collagen and inserted into a well in a 12-well plate. Triplicatewells were included per cell line and a well with no cells wasincluded as control. Once polymerized,mediawere placed on topof the collagen plug. After 24 hours, the collagen plugs weredetached from thewalls and bottomof thewell. Gelswere imaged(Syngene/G-Box) every 15minutes for thefirst fewhours followedby hourly images up to 24 hours. Gel area was measured bymanual outlining of the collagen plugs and analyzed in ImageJ.

Radioactive kinase assays[g-32P]-ATP (2.5 mCi; Perkin Elmer) suspended in kinase buffer

(27) was added to recombinant AURKA and/or CTTN (25–50 ng)protein and incubated for 30minutes at 30�C. Samples were thensubjected to gel electrophoresis and membrane transfer followedby X-ray film exposure for 2 hours to visualize phosphorylationand Western blotting to visualize total protein level.

Actin polymerization assaysActin polymerization assays were carried out according to the

manufacturer's recommendations (Cytoskeleton). RecombinantAURKA (3 mmol/L), CTTN (75 nmol/L), Arp2/3 (50 nmol/L,Cytoskeleton), or WASP-VCA domain (5 mmol/L, Cytoskeleton)suspended in magnesium/ATP cocktail were added to 5 mmol/Lpyrene-labeled G-actin in a 96-well plate and actin polymeriza-tion was measured on the basis of the increase of fluorescence atexcitation 365 nm/emission 407 nm every 30 seconds for 20minutes at 30�C using a GENios plate-reader (Tecan).

Cell kymographyKymography analysis was carried out as described previously

(18). In brief, cells were plated on glass bottom dishes (FisherScientific) 48 hours after nucleofection with mCherry-LifeAct.Cells were then treated overnight with 100 nmol/L AURKAinhibitor MLN8237 (Selleckchem) and live-imaged the next dayevery 45 seconds for 60minutes. Kymogramswere compiled froma perpendicular 1 pixel-width line from each frame on the samespot of the cell membrane.

Active Rac1 pulldown assaysCells were nucleofected with the indicated plasmids/siRNAs

and allowed to recover overnight, followed by suspension of1.5 � 105 cells in 450 mL of 1.5 mg/mL collagen-I or 4 mg/mLMatrigel. In indicated assays, cells were also treated overnightwith 4 mmol/L EHop-016 (ApexBio). After 24 hours, cells were

Targeting NEDD9/ROCK Inhibits TNBC Metastasis

www.aacrjournals.org Mol Cancer Res; 15(6) June 2017 671




lysed for 10 minutes on ice with 1� Rac1 lysis buffer asdescribed previously (28). Lysates were spun down at maxi-mum speed for 12 minutes at 4�C and supernatants were addedto Glutathione-4FastFlow sepharose (GE Healthcare) loadedwith GST or GST-PBD protein, rotating at 4�C for at least 1hour. Sepharose was then washed three times with lysis buffer,boiled in Laemmli buffer, and subjected to gel electrophoresisand Western blotting.

Active Rac1 and RhoA G-LISA assaysG-LISA activation assays for Rac1 and RhoA were performed as

recommended by the manufacturer's protocol (Cytoskeleton).

Immunoprecipitations of GEF proteinspcDNA3.1-mRFP-NEDD9 and Rac1 GEF-expressing plasmids

were cotransfected into 293T cells using TurboFect (ThermoFisher) as per the manufacturer's protocol. After 24 hours, cellswere lysed with PTY lysis buffer and processed as describedpreviously (27). Lysateswere preclearedwith protein-A/G-sephar-ose (GEHealthcare) for 1 hour at 4�C.Mouse IgG control/or anti-NEDD9 (2G9) protein-A/G sepharose, or anti-HA/anti-FLAGEZview-Red antibody (Sigma) were added to precleared lysatesfor 12 hours rotating at 4�C. Beads were then washed three timeswith PTY buffer. To elute proteins from the beads, 10% ammoniahydroxide was added for 15 seconds followed by speed-vacuumcentrifuge to remove ammonia hydroxide. Protein was thenresuspended in Laemmli buffer for gel electrophoresis and West-ern blotting. The immunoprecipitation (IP) of CTTN is detailed inthe Supplementary Materials and Methods.

Live-cell imaging of individual cell invasionCells treated with si/sgRNAs against NEDD9 or RhoA and/or

overexpressing CA-RhoA and/or treated with 50 mmol/L Y-27632or 100 nmol/L MLN8237 for 24 hours were resuspended in 1.5mg/mL collagen and placed into 3D-Chemotaxis m-slides (Ibidi)according to the manufacturer's protocol. FBS was added to theleft media reservoir to create a chemoattractant gradient. Whenused, Y-27632 and/or MLN8237 were also added to both reser-voirs to keep the drugs present during the assay. Cells were imagedevery 15 minutes via DIC 10� objective on a Nikon Sweptfield/Eclipse/TE2000-E confocal microscope for 23 hours. Time-lapseimages were imported into ImageJ software and combined toform video files. Quantification methods are detailed in theSupplementary Materials and Methods.

Live-cell imaging of collective cell invasionA total of 1�105 cells in fullmediumwere plated on topof 4–5

mg/mL Matrigel-coated delta-T dishes (Bioptechs). After 1 hour,the medium was removed and the cell layer was covered withmore Matrigel. After polymerizing for 30 minutes at 37�C, mediawas readded. Live-cell imaging was conducted every 10 minutesfor 24 hours at 37�C via Nikon Sweptfield/Eclipse/TE2000-Econfocal microscope (20� objective). Time-lapse images wereimported into ImageJ software and combined to form video files.Quantification methods are detailed in the Supplementary Mate-rials and Methods.

Fluorescent imaging of collective cell invasion sandwich assayCells were pretreated overnight with 50 mmol/L Y-27632 or

100 nmol/L MLN8237, and 3 � 104 cells were plated in 8-well

chamber slides (Thermo Fisher) coated with 150 mL 4 mg/mLMatrigel. After 1 hour, the medium was removed and the cellscovered with another 200 mL Matrigel, followed by full mediaafter Matrigel polymerization. Cells were live imaged in 5-mmZ-steps away from the cell seeding position at 72 hours using aNikon Sweptfield/Eclipse/TE2000-E confocal microscope. Inva-sion was quantified by total cell fluorescence intensity at thefurthest distance cells reached (100 mm). 3D invasion boxeswere constructed using the Volume-Viewer plugin for ImageJ toshow cell fluorescence signal throughout the Matrigel from theseeding position.

Mouse xenograft studyAnimals were housed in the West Virginia University (WVU)

Animal Facility under pathogen-free conditions according to anapproved Institutional Animal Care and Use Committee pro-tocol. A total of 1 � 106 MDA-MB-231-luc2 cells expressingdoxycycline-inducible pTRIPZ-RFP-shControl or shNEDD9(24) were suspended in 4 mg/mL Matrigel (Trevigen) andinjected into the mammary fat pad of 6- to 8-week-old immu-nodeficient female NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice(The Jackson Laboratory). Mice with mammary tumors (150–200 mm3) were pretreated in neoadjuvant settings with10mg/kg AURKA inhibitor MLN8237 via oral gavage for 1 week(4 days on, 3 days off), as described previously, to temporarilyhalt metastasis, but continue mammary tumor growth (30).shRNA expression in tumor cells was induced via introductionof doxycycline-containing food (Bio-Serv) in combination withvehicle or 10 mg/kg Y-27632 which was administered by oralgavage for 3 weeks (4 days on, 3 days off). Tumor growth anddissemination was evaluated weekly through in vivo whole-body bioluminescence imaging (BLI) using an IVIS/Lumina-IIsystem as described previously (16). After 3 weeks, lungs,mammary tumors, and blood were collected for analysis.Paraffin-embedded lung sections were stained by hematoxylinand eosin and analyzed for the number of metastases per lungarea by a pathologist as described previously (16). Primarytumors were analyzed for NEDD9 expression by Western blotanalysis. Ten mice per both shRNA groups (further separatedinto 5 per drug group for a total of 20 mice) were used based onstatistical analysis.

Quantification of circulating tumor cellsSubmandibular mouse blood samples were collected into

EDTA-coated tubes on ice to prevent clotting. Erythrocyteswere lysed and removed from blood via incubation with RBClysis buffer (eBioscience) according to the manufacturer'sprotocol. Cells were fixed in 2% paraformaldehyde for 10 min-utes, followed by centrifugation at 500 � g for 5 minutes at4�C, and resuspended in 1% BSA/PBS. FACS was performedusing a BD Biosciences/LSR Fortessa to count RFP-positivecirculating tumor cells (CTC) in the blood samples. Finalcounts were normalized to initial primary tumor size atweek zero.

Statistical analysisOne-way ANOVA or Student t test statistical analyses were

performed with GraphPad Prism software where noted. P �0.05 was considered to be significant (�). Experimental valueswere reported as the mean � SEM.

Jones et al.

Mol Cancer Res; 15(6) June 2017 Molecular Cancer Research672




ResultsTNBC cell morphology changes upon NEDD9 and ROCKinhibition

Metastatic TNBC cell lines with primarily mesenchymalmorphology (MDA-MB-231, BT549, HCC1143, HCC1395,Hs578T, and SUM159) were used to assess the role of NEDD9in switching cells from mesenchymal to amoeboid morphol-ogy. NEDD9 was depleted using previously characterizedsiRNAs (refs. 16, 31; Fig. 1A) in a SMARTpool, and the changesin individual cell morphology were monitored using live cellimaging microscopy when plated in 2D or suspended in 3Dmatrix (collagen I or Matrigel).

In 2D, control (siCon) cells have well-defined elongatedmorphology, while NEDD9-depleted (siNEDD9) cells show upto a 40% decrease in cell elongation (Supplementary Fig. S1Aand S1B) reflected by a decrease in the ratio of cell length overcell width. In 3D matrix, the pore size is a critical determinantof cell morphology and migration type (19). The average poresize of the collagen I matrix used was about 15 mm2 (Supple-mentary Fig. S1C and S1D), which supports both amoeboidand mesenchymal movement (32). Alternatively, the tightlypacked fibers that make up Matrigel are not conducive foramoeboid migration (<1 mm) and are permissive primarily formatrix-degrading mesenchymal movement. Interestingly, whenseeded in either of these 3D scaffolds, siNEDD9 cells showacquisition of rounded amoeboid morphology and up to a50% decrease in cell elongation (Fig. 1B and C), which issimilar to 2D results.

Knockout of NEDD9 using sgRNAs and CRISPR/Cas9 technol-ogy led to similar changes in morphology with less variability incell elongation and showing an elongation ratio closer to 1,indicating that they are more circular with similar cell length andwidth (Supplementary Fig. S1E and S1F). This phenotype wassimilar to cells overexpressing constitutively active RhoA GTPase(Fig. 1D–F), suggesting RhoA activity may be high in NEDD9-deficient cells. Importantly, almost all TNBC cells demonstratedsignificant plasticity in their ability to not only be able to switch toamoeboidmorphology, but also be able to increasemesenchymalmorphology, whichwas confirmed as control cells became furtherelongated upon application of the well-characterized Rho kinaseinhibitor, Y-27632 (Fig. 1G and H).

NEDD9 depletion leads to increased myosin light-chainphosphorylation but decreased contractility

Next, we tested whether the acquisition of amoeboid mor-phology by NEDD9-deficient TNBC cells is accompanied by aswitch to amoeboid migration, which could potentially pro-mote intra/extravasation during metastasis of breast cancer.Actomyosin contractility and force-mediated remodeling of thematrix is a signature of the amoeboid type of cell migration(33). A key driver of actomyosin contractility is the Rho kinaseROCK, which phosphorylates myosin II light chain (MLC2;ref. 34). In both collagen and Matrigel, NEDD9 deficiencyresulted in up to a threefold increase in pMLC2 (Fig. 2A–D;Supplementary Fig. S1J), which colocalized with cortical actinaround the cell membrane.

Interestingly, despite this increase in pMLC2 in NEDD9-depleted cells, there was a significant decrease in the abilityof the collective cell population to contract floating collagengels (Fig. 2E and F), indicating that the morphologically amoe-

boid NEDD9-depleted cells may not possess all of the mech-anistic properties of amoeboid cells. The inability of NEDD9-deficient cells to translate the increased pMLC2 signaling tomatrix deformation/contraction may be indicative of impairedadhesion or actin dynamics connecting the internal and exter-nal forces of the cell.

NEDD9 depletion leads to a decrease in mature focal adhesiondynamics

Mature, large focal adhesions have been associated with themesenchymal phenotype, whereas a decrease in size and matu-rity has been linked to amoeboid morphology (35, 36). Stain-ing for phospho-FAK (pFAKY397) or phospho-paxillin (pPax-illinY31) has been commonly used to label mature focal adhe-sions (36, 37). Immunofluorescence analysis shows up to a 2-fold decrease in the number of adhesions positive for pFAKY397

and pPaxillinY31 upon NEDD9 depletion (Fig. 2G–J; Supple-mentary Fig. S1G–S1J), indicating a critical role of NEDD9 inadhesion maturation. The decrease in FAK and paxillin phos-phorylation was further confirmed by Western blot analysis inmultiple breast cancer cell lines (Fig. 2K–M; Supplementary Fig.S2). NEDD9 deficiency trended toward a decrease in the num-ber of adhesions containing the more mature/stable focaladhesion marker vinculin (38), which would be concurrentwith amoeboid phenotype; however, the length of many ofthese adhesions was increased compared with control, suggest-ing the stabilization of fibrillar adhesions (Supplementary Fig.S3). Nevertheless, NEDD9-depleted cells also had an increasein the total number of small adhesions that contained talin(Supplementary Fig. S4), which is one of the first proteinsrecruited to focal contacts/adhesions (38). This indicates thatthere is an increase in small, weak nascent adhesions and adecrease in mature adhesions, which is concurrent with amoe-boid phenotype (39). This may suggest that NEDD9 is notneeded for adhesion initiation but is required for fibrillaryadhesion disassembly. Such duality may potentially affect bothmesenchymal and amoeboid adhesion dynamics.

NEDD9 regulates AURKA-driven phosphorylation of CTTN andstability of actin filaments

Amoeboid morphology is characterized by a decrease in fila-mentous actin polymerization while the mesenchymal elongatedshape relies on an increase infilamentous actin (7). Previously, wehave shown that NEDD9 binds and activates AURKA and theactin-stabilizing protein CTTN, which results in an increase inactin polymerization and stabilization of lamellar protrusionsand invadopodia (18). Here we show that AURKA directly phos-phorylates CTTN (Fig. 3A) at the (C-terminus 350–546 aa; Fig. 3B;Supplementary Fig. S5). To assess the impact of this phosphor-ylation on actin dynamics, an in vitro actin polymerization assaywas carried out using recombinant AURKA and CTTN proteins.Incubation of AURKAwith CTTN resulted in a significant increasein the actin polymerization rate (Fig. 3C) compared with CTTNwithout AURKA. AURKA alone or in combination with Arp2/3and WASP did not significantly affect actin polymerization, indi-cating that CTTN is the major substrate of AURKA in this assay.Moreover, kymography analysis of mCherry-LifeAct–expressingcells shows that the inhibition of AURKA results in a completeblockade of lamellar protrusions (Fig. 3D–F), indicating thatAURKA has a key role in cellular actin dynamics.






01

2

3456 P < 0.0001

P < 0.0001

P < 0.0001P < 0.0001

Cel

l elo

ngat

ion

(cel

l len

gth/

wid

th) *

* * * * ** *

* *

P < 0.0001

ns

HVehicle50 μmol/L Y-27632100 μmol/L Y-27632

MDA-MB-231 HCC1143 Hs578T BT549 HCC1395 SUM159

G HCC1395 SUM159BT549HCC1143 Hs578TMDA-MB-231

Vehi

cle

100

μmol

/L

Y-27

632

-50

-35-65

-50

+– +–+–CA-RhoA:

α-Tubulin

RhoA

BT549 HCC1395 SUM159

-115

-85-65

-50

+– +–+–siNEDD9:

α-Tubulin

NEDD9

F

E

0

1

2

33

siNEDD9siCon

CA-RhoA

BT549 HCC1395 SUM159

P < 0.0001 P < 0.0001 P < 0.0001

Cel

l elo

ngat

ion

(cel

l len

gth/

wid

th)

** * * * *HCC1395 SUM159BT549

siN

EDD

9C

A-R

hoA

siC

on

D

Cel

l elo

ngat

ion

(cel

l len

gth/

wid

th)

MDA-MB-231 HCC1143 Hs578TCol Mat Col Mat Col Mat

0

1

2

3

4

* ** * **

P = 0.04

P <

0.00

01

P = 0.03

P = 0.03

P <

0.00

01 P = 0.03

siNEDD9siConCHCC1143 Hs578TMDA-MB-231

siN

EDD

9si

Con

3D CollagenB

Hs578TsiNEDD9 #1 #2

siCon

NEDD9

GAPDH

-115

-80-40

-35

HCC1143

-115

-80-40

-35

siNEDD9 #1 #2

siCon

GAPDH

NEDD9

MDA-MB-231

-115

-80-40

-35

siNEDD9 #1 #2

siCon

NEDD9

GAPDH

A

Figure 1.

TNBC cell morphology changes upon NEDD9 and ROCK inhibition. A, Western blot analysis of NEDD9 expression in MDA-MB-231, HCC1143, andHs578T treated with multiple siRNAs. Bright-field images of MDA-MB-231, HCC1143, and Hs578T cells treated with siCon or siNEDD9 in 3D collagen I (B)and cell elongation quantified as cell length/width (40 cells/group; C). Bright-field images of BT549, HCC1395, and SUM159 cells expressing siCon,siNEDD9, or CA-RhoA (D), cell elongation quantified as cell length/width (100 cells/group; E), and Western blots of NEDD9 knockdown andCA-RhoA expression (F). Bright-field images of TNBC cells, vehicle, or Y-27632 treatment (G) and cell elongation quantified as cell length/width(40 cells/group; H). ns, not significant; � , P < 0.05, Student t test versus control.

Jones et al.





Ban

d in

tens

ity(%

to c

on)

siCon siNEDD9

pPaxillin

P = 0.0002

*

100

80

60

40

20

0

M

Ban

d in

tens

ity(%

to c

on)

0

20

40

60

80

100

*

siCon siNEDD9

pFAK

P = 0.0003

L

K

-40GAPDH

-115

-80NEDD9

-115FAK

-115pFAK

Paxillin

-65

-65

-50

-50

pPaxillin

siNEDD9 siConJ

0

25

50

75

100

*P < 0.0001

# fo

cal a

dhes

ions

/cel

l(%

to c

on)

siCon siNEDD9

RFI

(% to

con

)

I

siNEDD9siCon60

80

100

120

140

P = 0.002

*

pPaxillin

RFI

(% to

con

)

siNEDD9siCon0

50

100

150

200

*

pFAKH

P = 0.0002

siN

EDD

9si

Con

Merge/DNApFAKY397 Merge/DNApPaxillinY313D Collagen IG

0 1 2 3 4 5 60

25

50

75

100

23 2524

sgConsgNEDD9

Hours

Gel

are

a (%

to c

ontr

ol)

*P < 0.001

F Time: 0 1 2 3.5 24 h

Esg

NED

D9

sgC

onD

Ban

d in

tens

ity(%

to c

on)

0

50

100

150

200

250

*

siCon siNEDD9

P < 0.01

pMLC2

-25

-15MLC2

-25

-15pMLC2

siNEDD9 siCon

-40

-35GAPDH

-115

-80NEDD9

CpMLC2B

siCon siNEDD9

*300

200

P < 0.0001400

0

100RFI

(% to

con

)

A

siN

EDD

9

Merge/DNAActin

siC

on

pMLC2

Figure 2.

NEDD9 depletion increases phosphorylation of MLC2 and decreases phosphorylation of FAK and paxillin. A, MDA-MB-231 siCon and siNEDD9 cells incollagen I stained for pMLC2, actin, and Hoechst DNA dye. Scale bar, 20 mm. B, Box and whisker plot of pMLC2 relative fluorescence intensity (RFI), n ¼ 3, 20–30 cells/group. Intensity was normalized to the cell area. Western blot analysis (C) and quantification of MLC2 phosphorylation, n¼ 3 (D). Collagen gel contractionassay of MDA-MB-231 CRISPR sgCon/sgNEDD9 cells over 24 hours (E) and quantification of collagen gel area, n¼ 3 (F). G,MDA-MB-231 siCon and siNEDD9 cells incollagen I stained for pFAKY397 or pPaxillinY31 and Hoechst DNA dye. Scale bar, 20 mm. Box and whisker plot of pFAKY397 fluorescence (H) and pPaxillinY31

fluorescence, n ¼ 3, at least 20 cells/group (I). J, Quantification of # of focal adhesions per cell, n ¼ 3, at least 20 cells/group. K, Western blot analysis andquantification of FAK (L) and paxillin phosphorylation (M). n ¼ 3. �, P < 0.05, Student t test versus control.






NEDD9 depletion decreases Rac1 activity in breast cancercells

Rac1 GTPase has been shown to be one of the key drivers ofmesenchymal migration/morphology. To measure Rac1 activ-ity in cells grown in either 3D collagen or Matrigel, GST-taggedPBD (PAK p21–binding domain) protein was used to specif-ically pull down Rac1-GTP. Cells overexpressing dominant-negative Rac1 (DN) or constitutively active Rac1 (CA) wereused as controls (Supplementary Fig. S6A). NEDD9 depletionresulted in a significant decrease in the amount of active Rac1 incells grown in both scaffolds (Fig. 4A and B). This was consis-tent when repeated in two other mesenchymal breast cancer celllines as well, Hs578T and HCC1143 (Supplementary Fig. S6Band S6C). As a complementary approach, Rac1 G-LISA assays(40) were used to evaluate the effect of NEDD9 depletion.Similarly to the active Rac1 pulldown assay, Rac1 G-LISA showsa decrease in the amount of active Rac1 (Fig. 4B; SupplementaryFig. S6D). In addition, the RhoA G-LISA shows an increase inthe amount of active RhoA, which is concurrent with theamoeboid phenotype (Fig. 4C).

NEDD9 drives Rac1 activity through interaction with GEFVAV2

Previously, it was shown in melanoma cells that NEDD9activates Rac1 via interaction with the Rac1-GEF DOCK3 (11).On the basis of the breast cancer cell expression profiles of

different GEFs and their downstream effectors, several potentialRac1 GEF candidates were examined for their ability to bindNEDD9 in coimmunoprecipitation experiments includingDOCK4, DOCK180, VAV2, Tiam1, GEF-H1, ARHGEF2, ARH-GEF4, DEF6, ECT2, and DBL (Supplementary Table S1). SomeGEFs (such as DEF6, Supplementary Fig. S6E) were coimmu-noprecipitated with NEDD9 using anti-NEDD9–specific mono-clonal (2G9) antibody; however, only VAV2 was found to alsoreciprocally pull down NEDD9 (Fig. 4D; Supplementary Fig.S6F) when immunoprecipitating VAV2. To confirm that VAV2 isrequired for Rac1 activation in breast cancer cells, active Rac1pulldowns were performed on cells treated with siCon or siRNAtargeting VAV2 (Fig. 4E). VAV2 depletion resulted in a 40%–

60% reduction in the activity of Rac1 (Fig. 4F), supporting thenotion that VAV2 is a critical downstream effector of NEDD9-dependent Rac1 activation. Importantly, treatment of siConcells with EHop-016 (41), a small molecular compound whichinhibits Rac1 activation by VAV2 through interaction withRac1's VAV2-binding site, resulted in a reduction in Rac1 activitythat was similar to the effect caused by siNEDD9 (Supplemen-tary Fig. S6G and S6H). This inhibition effect by EHop-016 wasnot found to be additive when combined with NEDD9 deple-tion. In addition, an increase in Rac1 activity through NEDD9overexpression was able to be blocked by VAV2 knockdown,further supporting that NEDD9-induced Rac1 activity is VAV2dependent (Fig. 4G).

F

0

2

4

6

Veh MLN8237

Max

ext

ensi

on (

μm)

n.d.

Prot

rusi

on v

eloc

ity(n

m/s

ec)

E

0

5

10

15

20

Veh MLN8237

n.d.

Membrane protrusion

Tim

e

Vehicle MLN8237D

0 5 10 15 200

100

200

300

400

500

600

Arp2/3, WASP, CTTN, AURKAArp2/3, WASP, CTTNArp2/3, WASP, AURKAAURKA

G-Actin OnlyBuffer Only

*

Time (min)

Fluo

resc

ence

(AFU

)

CP < 0.0001

N-CTTN

C-CTTN

AURKA

-35-40

-65-50

-65-50

N-term CTTN:C-term CTTN:

AURKA:

--+

-+-

-++

+--

+-+

AURKAC-CTTN

-35-40

-65-50

B

P32

P32 -65-80

-50

AURKA

AURKA

CTTN

CTTN

AURKA:CTTN: -

++-

++

-50

-65-80

A

Figure 3.

NEDD9 regulates AURKA-drivenphosphorylation of CTTN and stabilityof actinfilaments.A,Radioactive in vitrokinase assay of recombinant AURKAand full-length wild-type (WT) CTTNproteins using radiolabeled P32-ATP, n¼ 3. Western blot analysis shows totalprotein (top and middle) whileautoradiograph shows P32-phosphorylated protein (bottom).B, Radioactive in vitro kinase assay ofrecombinant AURKA and N-term(1–350 aa) and C-term (350–546 aa)CTTN proteins using radiolabeledP32-ATP, n ¼ 3. Western blot analysisshows total protein (top and middle)while autoradiograph shows P32-phosphorylated protein (bottom,24-hour exposure). C, Pyrene-actinpolymerization assay of AURKA andCTTN, n ¼ 4. Curves were fit withBoltzmann sigmoidal analysis,� , P < 0.05. D, Kymographs of cellmembrane ofMDA-MB-231 cells treatedwith vehicle or MLN8237 andquantifications of protrusion velocity(E) andmaximummembrane extension(F). n.d., not detected.

Jones et al.





Simultaneous targeting of NEDD9 and ROCK/RhoA hindersbreast cancer cell motility/invasion in vitro

As NEDD9 depletion supports the ROCK/RhoA pathway,knockdown of NEDD9 was combined with the ROCK-targetingcompoundY-27632 to test for an additional therapeutic benefit in3D invasion assays. NEDD9 depletion alone (via siRNA orsgRNA) led to an increase in amoeboid morphology and drasti-

cally decreased invasion (Fig. 5A–E; Supplementary Videos S1 andS2; Supplementary Fig. S7A–S7D), similar to the overexpressionof constitutively active RhoA (Supplementary Fig. S7F–S7J; Sup-plementary Video S3). Combination of Y-27632 þ siNEDD9 ledto an increase in leading edge speed (extension/retraction),but decreased the cell body speed, thus rendering the elongatedcell immobile (Fig. 5C). The directionality and elongation of

-115

-15

-15

GST siCon siCon siVAV2

PD: Active

Rac1

TotalRac1

VAV2-115

NEDD9Exog.

Endog.

-140

-25

W

CL

G-25

α-Tub. -50

-80

GFP-NEDD9overexpression

Rac

1 A

ctiv

ity(%

to c

ontr

ol) P < 0.05

F

0

20

40

60

80

100

*

siCon siVAV2

E

W

CL

-15

-25

-25

-40 -80

-115-15

-25

GST siCon siVAV2

PD: ActiveRac1

Total Rac1

VAV2

GAPDH

D

-115

-80-140

-115

IgG NEDD9 IgG VAV2WCL

IP

VAV2

NEDD9

*

P =

0.0

27

3D Matrigel3D Collagen

Abs

orba

nce

(490

nm

)

2.0

1.5

1.0

0.5

0

G-LISA: Active RhoA

siConsiNEDD9

C

0.15-

0.10-

0.05-

0-

P =

0.0

36*

siCon siNEDD9

PD: Active Rac1

siConsiNEDD9

3D Collagen

0

20

40

60

80

100

*P = 0.02

siCon siNEDD9

3D Matrigel

Rac

1 A

ctiv

ity

(% to

con

trol

)

B

0

2040

60

80

100

*P < 0.05

W

CL-25Active

Rac1

Total Rac1

-15

GST siCon siNEDD9

PD:

A

-15

-25

-80

-115

-50

NEDD9

α-Tub.

Figure 4.

NEDD9 drives Rac1 activity through interaction with GEF VAV2. A, Western blot analysis of active-Rac1 pulldown (PD) from MDA-MB-231 siCon/siNEDD9 cellsin 3D Matrigel using GST-PBD–conjugated beads. Lane 1, GST-empty beads control. WCL, whole cell lysate. B, Quantification of Rac1 activity in 3D collagenand Matrigel (active/total Rac1) by Western blot analysis. C, Quantification of RhoA activity in 3D collagen and Matrigel by G-LISA. D, Coimmunoprecipitationof HA-VAV2 and RFP-NEDD9 through IP of each protein. IgG, mouse IgG control. Western blot analysis of active-Rac1 pulldown from MDA-MB-231 siCon/siVAV2cells (E) and quantification of Rac1 activity (F). G, Western blot analysis of active-Rac1 pulldown from GFP-NEDD9–overexpressing MDA-MB-231 siCon/siVAV2cells. Lane 1, GST-empty beads control. Endog, endogenous; Exog, exogenous; WCL, whole cell lysate. For all experiments: n ¼ 3. �, P < 0.05, Student t testversus control.






-115

-80-40

-35

shNEDD9 shCon

NEDD9

GAPDH

K

J

shNEDD9shCon

Inva

sion

(% to

con

)

0

25

50

75

100

*P = 0.006

Invasion Invasion

shCon shNEDD9I

Cel

l bod

y di

rect

iona

lity

(Euc

lidea

n/to

tal d

ista

nce)

0

0.2

0.4

0.6

0.8

1.0

*

shNEDD9shCon

P < 0.0001HG

shNEDD9shCon0

0.2

0.4

0.6

0.8

*

Cell

body

spe

ed (μ

m/m

in)

P < 0.0001

F shNEDD9shConGFP GFP

E

0

10

8

6

4

2Cel

l elo

ngat

ion

(cel

l len

gth/

wid

th)

Y-27632:siNEDD9: +

+-+

+-

--

P = 0.02 P = 0.005P < 0.0001

***C

ell b

ody

dire

ctio

nalit

y(E

uclid

ean/

tota

l dis

tanc

e)

Y-27632:siNEDD9: +

+-+

+-

--

0

0.2

0.4

0.6

0.8

1.0 * *P = 0.03P = 0.0005

DC

Cel

l spe

ed (μ

m/m

in)

LE CB LE CB LE CB LE CB

Y-27632:siNEDD9: +

+-+

+-

--

0

0.1

0.2

0.3

0.4

0.5

* * *ns

***P = 0.0001 P = 0.0003 P < 0.0001

P < 0.0001P = 0.008P = 0.01

-250-200-150-100-50 50

-100

-50

50

100

-250-200-150-100 -50 50

-100

-50

50

100

-250-200-150-100 -50 50

-100

-50

50

100

-250-200-150-100 -50 50

-100

-50

50

100

CH

EMO

ATT

RA

CTA

NT

BsiCon siNEDD9

Y-27

632

Vehi

cle

μm μm

μm

μm

AsiNEDD9siCon

Y-27

632

Vehi

cle

Figure 5.

Simultaneous targeting of NEDD9 and ROCK/RhoA hinders breast cancer cell motility/invasion in vitro. Bright-field outlined cell morphology images (�12 hours; A),individual cell tracking movement plots toward FBS chemoattractant (left; B), box and whisker plots of CB and LE speed (C), cell body directionality (D),and cell elongation of MDA-MB-231 cells with siCon/siNEDD9þ Vehicle/Y-27632 treatment in 3D collagen Ibidi chemotactic invasion movies (E). Scale bar, 20 mm.n¼ 3, 30–40 cells per group; ns, not significant; � , P < 0.05, one-way ANOVA. GFP immunofluorescence images and box and whisker plots (F) of cell body speed (G)and cell body directionality of MDA-MB-231 pGIPZ-GFP shControl and shNEDD9 cells after 24 hours being densely seeded in a 3D Matrigel sandwich assay (H).Scale bar, 20 mm. n ¼ 3, at least 60 cells per group; � , P < 0.05, Student t test versus control. 3D projection boxes (I) and collective invasion quantification (J)of the amount of MDA-MB-231 pGIPZ-GFP shControl/shNEDD9 cells reaching 100-mm invasion distance after 72 hours, being densely seeded in a 3D Matrigelsandwich assay. �, P < 0.05, Student t test versus control. K, Western blot analysis of shNEDD9 knockdown in a collective invasion assay.

Jones et al.





Y-27632–treated cells was significantly increased regardless ofNEDD9depletion (Fig. 5DandE). As a complementary approach,RhoAdepletionwas performed and resulted in comparable trends(Supplementary Fig. S8; Supplementary Video S4). Effects oninvasion were also evaluated using Hs578T and HCC1143 cellstreatedwith siNEDD9, confirming similar results (SupplementaryFig. S9; Supplementary Videos S5 and S6). In addition, theAURKA inhibitor MLN8237 was tested as a substitute forsiNEDD9 in combination treatments. Cell invasion distance wasfound to be decreased with MLN8237/Y-27632 treatment, andwas similar to the efficacy of siNEDD9/Y-27632 treatment (Sup-plementary Fig. S10A and S10B; Supplementary Video S7). Thissuggests that MLN8237/Y-27632 could potentially be a viablecombination regimen as well.

NEDD9 is required for collective migration in 3D matrixTo test the impact of NEDD9 depletion on collective tumor cell

invasion, cells were seeded in high density while sandwiched inbetween two Matrigel layers and time-lapse microscopy wasperformed. Control cells invaded as strand-like collective streams("highways"), similar to previous observations (ref. 9; Fig. 5F;Supplementary Video S8). Formation of these streams is dimin-ished following NEDD9 knockdown, which suggests the cellshaveminimalmatrix degradation and/or rearrangement capacity.This deficiency in path-generating activity is required for leadingcells during collective invasion. In addition, cell speed and direc-tionality were both significantly decreased in NEDD9-depletedcells (Fig. 5G and H), which corroborates with a significantreduction in their invasion distance (Fig. 5I–K). Furthermore,inhibition of AURKA yielded a similar decrease in collective cellinvasion compared with NEDD9 depletion (Supplementary Fig.S10C and S10D), again suggesting that AURKA could be a viabletherapeutic target.

Combination of NEDD9 knockdown and Y-27632 treatmenthinders breast cancer cell invasion and metastasis in vivo

MDA-MB-231-luc2 pTRIPZ-RFP-shControl or shNEDD9 cellswere injected into the mammary fat pad of NSG mice. After 2.5weeks, mice were treated with the AURKA inhibitor MLN8237 (4days on, 3 days off) to prevent pulmonary metastases outgrowthwithout any effect on themammary tumor growth as we reportedpreviously (16). Seventy-two hours post-MLN8237 pretreatment,mice with tumors were subjected to doxycycline-induced shConor shNEDD9 expression and administration of Y-27632 com-pound for 3weeks (Fig. 6A). Primary tumor growthwasmeasuredweekly and was significantly reduced in shNEDD9/Y-27632 com-bination-treated animals compared with vehicle (Fig. 6B and C).At the end of the study, NEDD9 depletion was confirmed in themammary tumors by Western blot analysis (Fig. 6D and E). Tomeasure the metastatic ability of the tumor cells, terminal bloodsamples were collected and analyzed after euthanization forCTCs using flow cytometry. Nearly a 50% reduction in thenumber of CTCs was documented with shNEDD9 induction,which decreased even further with the addition of Y-27632(Fig. 6F). A significant abatement in the number of metastasesper lung area was seen with NEDD9 knockdown, and an evenmore pronounced inhibition was seen in combination withY-27632 (Fig. 6G), supporting our hypothesis that dual targetingof both the mesenchymal and amoeboid pathways via NEDD9andROCK/RhoA can be a promising therapy against breast cancermetastasis (Fig. 6H).

DiscussionNEDD9 protein is found consistently upregulated in TNBCs

and highly correlates with invasive/metastatic spread (1, 16). Ourfindings provide a mechanistic explanation for NEDD9-driveninvasion processes and strongly advocates for the combination ofanti-NEDD9/AURKA and anti-ROCK–targeting compounds toinhibit these movement signaling cascades. In this study, wereport that deficiency in NEDD9 signaling itself leads to inhibi-tion of key aspects of both mesenchymal and amoeboid migra-tion in TNBC cells, resulting in substantial hindrance on cellinvasion and metastasis. Similar to results in melanoma (14),NEDD9 deficiency in TNBC cells results in rounded/amoeboidmorphology along with a decrease in the total number of mature(pFAK/pPaxillin positive) adhesions and an increase in the num-ber of nascent adhesions (39). However, these cells also saw anincrease in long fibrillar adhesions (Supplementary Fig. S3D).These findings suggest that NEDD9 is also required for thedisassembly of fibrillar adhesions similar to vinculin (42), whichregulates the recruitment and release of focal adhesion proteins ina force-dependent manner (43). The role of NEDD9/HEF1 as asensor of altered adhesion states has been previously reported(44). A decrease in disassembly could be due to matrix metallo-protease inhibition (45), as previously reported by our group(23).Hindered disassembly of FAswouldbe indisagreementwitha previously noted increase in adhesion turnover of NEDD9-KOor mutant NEDD9 (Y189A) expressing fibroblasts in 2D cultures(46, 47); however the structure, composition, and dynamics ofadhesions in 2D fibroblastsmay be different from 3D tumor cells.

NEDD9 deficiency led to an increase in myosin-phosphorylat-ed amoeboid-like cells, but this increase in pMLC2 did nottranslate to increased collective-cell contractility or invasion in3D matrix. This disconnect may suggest a potential decouplingbetween the actin filaments and myosin motors during cell bodycontraction (48). Alternatively, the inability of NEDD9-deficientcells to disassemble adhesions as we have suggested would limitamoeboid movement, which is in agreement with our previousreports on integrin dynamics (31).

Consistent with pastfindings, we found thatNEDD9deficiencyresulted in over a 40% decrease in active Rac1, suggesting thatNEDD9 is critical for Rac1 activation to occur in TNBC cells. Wediscovered that an interaction betweenNEDD9 and the Rac1-GEFprotein VAV2 is required for Rac1 activation. Similar to NEDD9depletion, loss of VAV2 yields a comparable decrease in Rac1activity. Furthermore, the small-molecule compound EHop-016,which efficiently blocks the interaction of Rac1 with VAV2 (41),did not cause an additive effect on Rac1 inhibition when given tosiNEDD9 cells, suggesting that NEDD9 and VAV2 share the samepathway to Rac1 activation. The potential clinical application ofthis compound has yet to be determined; however, another VAV-family–targeting drug, the purine analogue azathioprine, wasrecently found to inhibit pancreatic cancer metastasis (49). Inaddition, overexpression of NEDD9 increased Rac1 activity,which was abrogated upon depletion of VAV2, supporting thatNEDD9's influence onRac1 activation is VAV2-dependent. Select-ing NEDD9 as a therapeutic target could potentially be morebeneficial than targeting VAV2 due to its upstream placement inthe mesenchymal pathway and the multiple downstreambranches that it links to including VAV2/Rac1 and AURKA/CTTN.The effect of AURKA inhibition on actin dynamics in TNBCs andits ability to regulate filamentous actin polymerization via






phosphorylation of CTTN suggests a potential mechanistic expla-nation for the mesenchymal to amoeboid morphology changesobserved uponNEDD9depletion andprovides additional venuesto explore CTTN/AURKA signaling in TNBC metastasis.

In collagen invasion assays, both NEDD9-deficient andMLN8237-treated cells underwent a drastic reduction in cellinvasion. It was anticipated that amoeboid cells could potentiallymove faster with a more erratic cell trajectory. While a smalldecrease in directionality was seen uponNEDD9depletion, it wasnot significant. Inhibition of ROCK or RhoA in NEDD9-deficientTNBC cells reverted them back to an elongated cell shape, but nottheir original migration proficiency. These combination-treatedcells also had a significant increase in cell directionality—this islikely due to their significant increase in cell lengthwhichwas evenhigher than control cells.

The substantially impeded movement of combination-treatedcells could be related to the large difference in speed observedbetween the cell body (CB) and leading edge (LE). LE speedremains high, continuing to protrude and retract, but the cell isincompetent todegradematrix and/ormake anchor points, due toNEDD9 deficiency. Meanwhile, the cell rear cannot efficientlycontract due to inhibition of ROCK/RhoA, leaving the CB moreimmobile. This combination therefore results in a severe gapbetween a stagnant, immotile CB along with a quickly protrud-ing/retracting, yet functionally impaired LE. These results supportthe notion that simultaneous inhibition of both the NEDD9/AURKA and ROCK/RhoA pathways could be an efficient antimi-gratory/metastasis treatment option. These conclusions were fur-ther supported by our findings in an in vivo breast cancer xenograftmouse model, demonstrating a significant reduction in the

Mesenchymal Amoeboid

VAV2

RAC1

WASP Arp2/3

AURKA

CTTN

NEDD9

P

Actin polymerization andlamellipodia/invadopodia formation

ROCK

RHOA

MLCP

Actomyosincontractility

H

# m

ets/

mm

2(%

to c

on)

* *

*

P =

0.0

02

P =

0.0

09

P =

0.0

4

100

75

50

250

G

Y-27632:shNEDD9: +

+–+

+–

––

F

# C

TCs

(% to

con

)

P <

0.0

001

P <

0.0

001125

10075

50

25

0

**

Y-27632:shNEDD9: +

+–+

+–

––

P = 0.006

shCon shNEDD9

*100

50

E

0NED

D9

Expr

essi

on

(% to

con

)

-115

-80

-50α-Tub.

NEDD9

Y-27632:shNEDD9: +

+–+

+–

––

D

Time of treatment (weeks)

shNEDD9+VehicleshCon+Vehicle

shCon+Y-27632shNEDD9+Y-27632

150

1 2 30

50

100 *

C

Prim

ary

tum

or g

row

th

(pho

tons

/sec

/cm

2 /sr,

% to

wee

k 0)

B

shCon+ Vehicle

shNEDD9+ Vehicle

shCon+ Y-27632

shNEDD9+ Y-27632

0 1 2 3

8

6

4

2

0

Radiance (×10 8, photons/sec/cm

2/sr)

Time of treatment (weeks)

Week: –3.5 –1 0 1 2 3

Inject cancer cells MLN8237 10 mg/kg

4 days on, 3 days offVehicle/Y-27632 10 mg/kg treatment

Doxycycline food (shCon/shNEDD9 induction)+

A

Figure 6.

Simultaneous targeting of NEDD9 andROCK hinders breast cancer cellinvasion andmetastasis in vivo.A,MDA-MB-231-luc2 TZ RFP-shCon and RFP-shNEDD9 cells injected into themammary fat pad of NSG mice weregrown for approximately 3.5 weeksbefore starting treatment (time 0) for3 weeks with 10 mg/kg Y-27632 andshRNA induction. B, Weekly images oftumor cells in vivo via BLI signal whichwas quantified at the primary tumor site(C). Western blot analysis (D) andquantification (E) of NEDD9knockdown in final primary tumors.F, Quantification of number of RFP-positive circulating tumor cells inmouse blood by FACS analysis.G, Quantification of number of visiblemetastases per area of lung tissue.H, Proposed schematic ofmesenchymal and amoeboid pathwaysin breast cancer. n ¼ 5 mice/group;� , P < 0.05, one-way ANOVA.

Jones et al.





number of circulating tumor cells and pulmonary metastases ofmice that received combination treatment. These findings are alsoin agreement with previously published reports of NEDD9 deple-tion showing a decrease in the invasion and dissemination ofdiverse cancer types in vivo (1, 2).

In conclusion, while NEDD9-deficient cells may have manyhallmarks of amoeboid phenotype, we have shown that amoe-boid movement appears defective in some respects such asdecreased cell contractility. Nonetheless, depletion of RhoA oraddition of ROCK inhibitor to NEDD9-deficient cells providedadditional benefit in further hindrance of cell invasion/metasta-sis. Through concurrent targeting of these pathways, this approachmerits further evaluation as a clinical option to augment breastcancer patient survival.

Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.

DisclaimerThis manuscript contains original work only and has not been published or

submitted elsewhere. All of the authors have directly participated in theplanning, execution and/or analysis of this study and have approved thesubmitted version of this manuscript.

Authors' ContributionsConception and design: B.C. Jones, L.C. Kelley, E.N. PugachevaDevelopment of methodology: B.C. Jones, L.C. Kelley, Y.V. Loskutov

Acquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): B.C. Jones, L.C. Kelley, Y.V. Loskutov, K.M. Marinak,V.K. Kozyreva, E.N. PugachevaAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): B.C. Jones, V.K. Kozyreva, M.B. Smolkin, E.N.PugachevaWriting, review, and/or revision of the manuscript: B.C. Jones, Y.V. Loskutov,E.N. PugachevaAdministrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): B.C. Jones, E.N. PugachevaStudy supervision: E.N. PugachevaOther (pathology/histologic analysis of tissue samples): M.B. Smolkin

AcknowledgmentsThe authors would like to thank and acknowledge theWVU Shared Research

Facilities and theMBRCCandHSCcore facilities including theWVUMicroscopeImaging Facility and the WVU Flow Cytometry Core Facility.

Grant SupportThis research was supported by grants CA148671 (to E.N. Pugacheva) and

KG100539 (to E.N. Pugacheva) from the NIH/NCI and Susan G. Komen forthe Cure foundation, and in part by a NIH/NCRR5 P20-RR016440-09.The WVU HSC Core Facilities were supported by the NIH grants P30-RR032138/GM103488, S10-RR026378, S10-RR020866, S10-OD016165,and P20GM103434.

The costs of publication of this articlewere defrayed inpart by the payment ofpage charges. This article must therefore be hereby marked advertisement inaccordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Received November 12, 2016; revised December 14, 2016; accepted February4, 2017; published OnlineFirst February 24, 2017.

References1. Kong C, Wang C, Wang L, Ma M, Niu C, Sun X, et al. NEDD9 is a positive

regulator of epithelial-mesenchymal transition and promotes invasion inaggressive breast cancer. PLoS One 2011;6:e22666.

2. Kim M, Gans JD, Nogueira C, Wang A, Paik JH, Feng B, et al. Comparativeoncogenomics identifies NEDD9 as a melanoma metastasis gene. Cell2006;125:1269–81.

3. Friedl P, Wolf K. Plasticity of cell migration: a multiscale tuning model.J Cell Biol 2010;188:11–9.

4. Friedl P, Wolf K. Tumour-cell invasion andmigration: diversity and escapemechanisms. Nat Rev Cancer 2003;3:362–74.

5. Haeger A, Wolf K, Zegers MM, Friedl P. Collective cell migration: guidanceprinciples and hierarchies. Trends Cell Biol 2015;25:556–566.

6. Yilmaz M, Christofori G. Mechanisms of motility in metastasizing cells.Mol Cancer Res 2010;8:629–42.

7. Pankova K, Rosel D, Novotny M, Brabek J. The molecular mechanisms oftransition between mesenchymal and amoeboid invasiveness in tumorcells. Cell Mol Life Sci 2010;67:63–71.

8. Wolf K, Friedl P. Extracellular matrix determinants of proteolytic and non-proteolytic cell migration. Trends Cell Biol 2011;21:736–44.

9. Wolf K, Mazo I, Leung H, Engelke K, von Andrian UH, Deryugina EI, et al.Compensation mechanism in tumor cell migration: mesenchymal-amoe-boid transition after blocking of pericellular proteolysis. J Cell Biol2003;160:267–77.

10. Case LB, BairdMA, Shtengel G, Campbell SL,HessHF,DavidsonMW, et al.Molecular mechanism of vinculin activation and nanoscale spatial orga-nization in focal adhesions. Nat Cell Biol 2015;17:880–92.

11. Ahn J, Sanz-Moreno V, Marshall CJ. The metastasis gene NEDD9product acts through integrin beta3 and Src to promote mesenchymalmotility and inhibit amoeboid motility. J Cell Sci 2012;125(Pt 7):1814–26.

12. Bradbury P, Mahmassani M, Zhong J, Turner K, Paul A, Verrills NM, et al.PP2A phosphatase suppresses function of the mesenchymal invasionregulator NEDD9. Biochim Biophys Acta.1823:290–7.

13. Parri M, Chiarugi P. Rac and Rho GTPases in cancer cell motility control.Cell Commun Signal 2010;8:23.

14. Sanz-Moreno V, Gadea G, Ahn J, Paterson H, Marra P, Pinner S, et al. Racactivation and inactivation control plasticity of tumor cell movement. Cell2008;135:510–23.

15. AmmerAG,Weed SA.Cortactin branches out: roles in regulating protrusiveactin dynamics. Cell Motil Cytoskeleton 2008;65:687–707.

16. Ice RJ, McLaughlin SL, Livengood RH, Culp MV, Eddy ER, Ivanov AV, et al.NEDD9 depletion destabilizes Aurora A kinase and heightens the efficacyof Aurora A inhibitors: implications for treatment of metastatic solidtumors. Cancer Res 2013;73:3168–80.

17. Pugacheva EN, Jablonski SA, Hartman TR, Henske EP, Golemis EA. HEF1-dependent Aurora A activation induces disassembly of the primary cilium.Cell 2007;129:1351–63.

18. Kozyreva VK, McLaughlin SL, Livengood RH, Calkins RA, Kelley LC,Rajulapati A, et al. NEDD9 regulates actin dynamics through cortactindeacetylation in an AURKA/HDAC6-dependent manner. Mol Cancer Res2014;12:681–93.

19. Wolf K, Te Lindert M, Krause M, Alexander S, Te Riet J, Willis AL, et al.Physical limits of cell migration: control by ECM space and nucleardeformation and tuning by proteolysis and traction force. J Cell Biol2013;201:1069–84.

20. Friedl P, Wolf K, Lammerding J. Nuclear mechanics during cell migration.Curr Opin Cell Biol 2011;23:55–64.

21. Seo S, Ichikawa M, Kurokawa M. Structure and function of cas-L andintegrin-mediated signaling. Crit Rev Immunol 2006;26:391–406.

22. Aquino JB, Lallemend F,Marmigere F, Adameyko II, Golemis EA, Ernfors P.The retinoic acid inducible Cas-family signaling protein Nedd9 regulatesneural crest cell migration by modulating adhesion and actin dynamics.Neuroscience 2009;162:1106–19.

23. Loskutov YV, Kozyulina PY, Kozyreva VK, Ice RJ, Jones BC, Roston TJ, et al.NEDD9/Arf6-dependent endocytic trafficking of matrix metalloproteinase14: a novel mechanism for blocking mesenchymal cell invasion andmetastasis of breast cancer. Oncogene 2014;34:3662–75.

24. McLaughlin SL, Ice RJ, Rajulapati A, Kozyulina PY, LivengoodRH,KozyrevaVK, et al. NEDD9 depletion leads to MMP14 inactivation by TIMP2 andprevents invasion and metastasis. Mol Cancer Res 2014;12:69–81.






25. Yamazaki D, Kurisu S, Takenawa T. Involvement of Rac and Rho signalingin cancer cell motility in 3D substrates. Oncogene 2009;28:1570–83.

26. Weaver AM, Karginov AV, Kinley AW, Weed SA, Li Y, Parsons JT, et al.Cortactin promotes and stabilizes Arp2/3-induced actin filament networkformation. Curr Biol 2001;11:370–4.

27. Pugacheva EN, Golemis EA. The focal adhesion scaffolding protein HEF1regulates activation of the Aurora-A and Nek2 kinases at the centrosome.Nat Cell Biol 2005;7:937–46.

28. Wozniak MA, Keely PJ. Use of three-dimensional collagen gels to studymechanotransduction in T47D breast epithelial cells. Biol Proced Online2005;7:144–61.

29. Castello-Cros R, Cukierman E. Stromagenesis during tumorigenesis: char-acterization of tumor-associated fibroblasts and stroma-derived 3Dmatri-ces. Methods Mol Biol 2009;522:275–305.

30. Kozyreva VK, Kiseleva AA, Ice RJ, Jones BC, Loskutov YV, Matalkah F, et al.Combination of Eribulin and Aurora A inhibitor MLN8237 preventsmetastatic colonization and induces cytotoxic autophagy in breast cancer.Mol Cancer Ther 2016;15:1809–22.

31. Kozyulina PY, Loskutov YV, Kozyreva VK, Rajulapati A, Ice RJ, Jones BC,et al. Prometastatic NEDD9 regulates individual cell migration via caveo-lin-1-dependent trafficking of integrins. Mol Cancer Res 2015;13:423–38.

32. Artym VV, Matsumoto K. Imaging cells in three-dimensional collagenmatrix. Curr Protoc Cell Biol 2010;Chapter 10:Unit 10 81–20.

33. Gaggioli C, Sahai E. Melanoma invasion - current knowledge and futuredirections. Pigment Cell Res 2007;20:161–72.

34. Wyckoff JB, Pinner SE, Gschmeissner S, Condeelis JS, Sahai E. ROCK- andmyosin-dependent matrix deformation enables protease-independenttumor-cell invasion invivo. Curr Biol 2006;16:1515–23.

35. Carragher NO, Walker SM, Scott Carragher LA, Harris F, Sawyer TK,Brunton VG, et al. Calpain 2 and Src dependence distinguishes mesen-chymal and amoeboid modes of tumour cell invasion: a link to integrinfunction. Oncogene 2006;25:5726–40.

36. Hager MH,Morley S, Bielenberg DR, Gao S, MorelloM, Holcomb IN, et al.DIAPH3 governs the cellular transition to the amoeboid tumour pheno-type. EMBO Mol Med 2012;4:743–60.

37. Cukierman E, Pankov R, Stevens DR, Yamada KM. Taking cell-matrixadhesions to the third dimension. Science 2001;294:1708–12.

38. Wozniak MA, Modzelewska K, Kwong L, Keely PJ. Focal adhesion regula-tion of cell behavior. Biochim Biophys Acta 2004;1692:103–19.

39. Huttenlocher A, Horwitz AR. Integrins in cell migration. Cold Spring HarbPerspect Biol 2011;3:a005074.

40. Schlegel N, Burger S, Golenhofen N, Walter U, Drenckhahn D, Waschke J.The role of VASP in regulation of cAMP- and Rac 1-mediated endothelialbarrier stabilization. Am J Physiol 2008;294:C178–88.

41. Montalvo-Ortiz BL, Castillo-Pichardo L, Hernandez E, Humphries-BickleyT, De laMota-Peynado A, Cubano LA, et al. Characterization of EHop-016,novel small molecule inhibitor of Rac GTPase. J Biol Chem 2012;287:13228–38.

42. Teckchandani A, Cooper JA. The ubiquitin-proteasome system regulatesfocal adhesions at the leading edge of migrating cells. eLife 2016;5.pii:e17440.

43. Carisey A, Tsang R, Greiner AM, Nijenhuis N, Heath N, Nazgiewicz A, et al.Vinculin regulates the recruitment and release of core focal adhesionproteins in a force-dependent manner. Curr Biol 2013;23:271–81.

44. O'Neill GM, Golemis EA. Proteolysis of the docking protein HEF1 andimplications for focal adhesion dynamics. Mol Cell Biol 2001;21:5094–108.

45. Stehbens SJ, PaszekM, PembleH, Ettinger A,Gierke S,Wittmann T. CLASPslink focal-adhesion-associatedmicrotubule capture to localized exocytosisand adhesion site turnover. Nat Cell Biol 2014;16:561–73.

46. Zhong J, Baquiran JB, Bonakdar N, Lees J, Ching YW, Pugacheva E, et al.NEDD9 stabilizes focal adhesions, increases binding to the extra-cellularmatrix and differentially effects 2D versus 3D cell migration. PLoS One2012;7:e35058.

47. Zou Z, Yuan Z, Zhang Q, Long Z, Chen J, Tang Z, et al. Aurora kinase Ainhibition-induced autophagy triggers drug resistance in breast cancer cells.Autophagy 2012;8:1798–810.

48. Wollrab V, Thiagarajan R, Wald A, Kruse K, Riveline D. Still and rotatingmyosin clusters determine cytokinetic ring constriction. Nat Commun2016;7:11860.

49. Razidlo GL, Magnine C, Sletten AC, Hurley RM, Almada LL,Fernandez-Zapico ME, et al. Targeting pancreatic cancer metastasisby inhibition of Vav1, a driver of tumor cell invasion. Cancer Res2015;75:2907–15.


Jones et al.




2017;15:670-682. Published OnlineFirst February 24, 2017.Mol Cancer Res Brandon C. Jones, Laura C. Kelley, Yuriy V. Loskutov, et al. Metastatic BehaviorDual Targeting of Mesenchymal and Amoeboid Motility Hinders

Updated version

10.1158/1541-7786.MCR-16-0411doi:

Access the most recent version of this article at:

Material

Supplementary

http://mcr.aacrjournals.org/content/suppl/2017/02/23/1541-7786.MCR-16-0411.DC1

Access the most recent supplemental material at:

Cited articles

http://mcr.aacrjournals.org/content/15/6/670.full#ref-list-1

This article cites 46 articles, 16 of which you can access for free at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://mcr.aacrjournals.org/content/15/6/670To request permission to re-use all or part of this article, use this link



http://mcr.aacrjournals.org/lookup/doi/10.1158/1541-7786.MCR-16-0411

http://mcr.aacrjournals.org/content/suppl/2017/02/23/1541-7786.MCR-16-0411.DC1

http://mcr.aacrjournals.org/content/15/6/670.full#ref-list-1

http://mcr.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://mcr.aacrjournals.org/content/15/6/670


dual targeting of mesenchymal and amoeboid...

Documents