Retraction

CELL BIOLOGYRetraction for “Inverted formin 2 in focal adhesions promotesdorsal stress fiber and fibrillar adhesion formation to drive extra-cellular matrix assembly,” by Colleen T. Skau, Sergey V. Plotnikov,Andrew D. Doyle, and Clare M. Waterman, which was first pub-lished April 27, 2015; 10.1073/pnas.1505035112 (Proc Natl Acad SciUSA 112:E2447–E2456).The authors wish to note the following: “The US Public Health

Service (PHS) has made a finding of research misconduct againstColleen T. Skau. Based on Dr. Skau’s admission, an assessmentconducted by NIH, and analysis conducted by the Office ofResearch Integrity (ORI) in its oversight review, ORI found thatDr. Skau engaged in research misconduct in research supportedby National Heart, Lung, and Blood Institute, NIH. The fol-lowing is a list of falsified data published in the article:

1. Figs. 1E, 2A, 5B, and 5C were falsified by selectively includingand/or omitting data points from the analyses.

2. Figs. 1E (Top), 2C (Middle and Right), 3A (Right), 3C (Right),5B, 5C, and 7D (Right) were falsified by reporting more datapoints than were actually measured.

3. Figs. 2A and 3C (Right) were falsified by reporting data thatdid not originate from genuine experimental observations.

4. Fig. 2C (Left and Right), 3A (Right), 5B, and 5C were falsifiedby performing statistical calculations and reporting signifi-cance based on falsified data.

5. Fig. 2C (Left and Right) and 5B were falsified by selectivelymanipulating raw measurements.

6. Fig. 7A was falsified by reporting that error bars representedSD, when they actually represented SEM.

“The first author, Colleen T. Skau, is solely responsible forthese actions. All authors apologize to the scientific communityfor any inconvenience this might have caused. We hereby retractthis article.”

Published under the PNAS license.

Published online March 12, 2018.

www.pnas.org/cgi/doi/10.1073/pnas.1803125115

E2900 | PNAS | March 20, 2018 | vol. 115 | no. 12 www.pnas.org

Dow

nloa

ded

by g

uest

on

Oct

ober

18,

202

0 D

ownl

oade

d by

gue

st o

n O

ctob

er 1

8, 2

020

Dow

nloa

ded

by g

uest

on

Oct

ober

18,

202

0 D

ownl

oade

d by

gue

st o

n O

ctob

er 1

8, 2

020

Dow

nloa

ded

by g

uest

on

Oct

ober

18,

202

0 D

ownl

oade

d by

gue

st o

n O

ctob

er 1

8, 2

020

Dow

nloa

ded

by g

uest

on

Oct

ober

18,

202

0 D

ownl

oade

d by

gue

st o

n O

ctob

er 1

8, 2

020

Dow

nloa

ded

by g

uest

on

Oct

ober

18,

202

0 D

ownl

oade

d by

gue

st o

n O

ctob

er 1

8, 2

020

Dow

nloa

ded

by g

uest

on

Oct

ober

18,

202

0 D

ownl

oade

d by

gue

st o

n O

ctob

er 1

8, 2

020

Inverted formin 2 in focal adhesions promotes dorsalstress fiber and fibrillar adhesion formation to driveextracellular matrix assemblyColleen T. Skaua, Sergey V. Plotnikovb, Andrew D. Doylec, and Clare M. Watermana,1

aCell Biology and Physiology Center, National Heart Lung and Blood Institute, and cCell Biology Section, Laboratory of Cell and Developmental Biology,National Institute of Dental and Craniofacial Research, NIH, Bethesda, MD 20892; and bDepartment of Cell and Systems Biology, University of Toronto,Toronto, ON, Canada M5S 3G5

Edited by Martin A. Schwartz, Yale School of Medicine, New Haven, CT, and accepted by the Editorial Board March 31, 2015 (received for review March18, 2015)

Actin filaments and integrin-based focal adhesions (FAs) formintegrated systems that mediate dynamic cell interactions withtheir environment or other cells during migration, the immuneresponse, and tissue morphogenesis. How adhesion-associatedactin structures obtain their functional specificity is unclear. Herewe show that the formin-family actin nucleator, inverted formin 2(INF2), localizes specifically to FAs and dorsal stress fibers (SFs) infibroblasts. High-resolution fluorescence microscopy and ma-nipulation of INF2 levels in cells indicate that INF2 plays a criticalrole at the SF–FA junction by promoting actin polymerization viafree barbed end generation and centripetal elongation of an FA-associated actin bundle to form dorsal SF. INF2 assembles into FAsduring maturation rather than during their initial generation, andonce there, acts to promote rapid FA elongation and maturationinto tensin-containing fibrillar FAs in the cell center. We show thatINF2 is required for fibroblasts to organize fibronectin into matrixfibers and ultimately 3D matrices. Collectively our results indicatean important role for the formin INF2 in specifying the function offibrillar FAs through its ability to generate dorsal SFs. Thus, dorsalSFs and fibrillar FAs form a specific class of integrated adhesion-associated actin structure in fibroblasts that mediates generationand remodeling of ECM.

INF2 | integrin | actin | fluorescence microscopy

The dynamic connection between the forces generated in theactomyosin cytoskeleton and integrin-mediated focal adhe-

sions (FAs) to the extracellular matrix (ECM) is essential formany physiological processes including cell migration, vascularformation and function, the immune response, and tissue mor-phogenesis. These diverse functions are mediated by distinctcellular structures including protruding lamellipodia containingnascent FAs that mediate haptotaxis (1), ventral adhesive actinwaves that mediate leukocyte transmigration through endothelia(2, 3), and stress fibers (SFs) and FAs that drive fibrillarization ofECM in developing embryos (4, 5). The coordination and in-terdependence of actin and integrin-based adhesion in thesespecialized cellular structures are rooted in their biochemicalinterdependence. Activation of integrins to their high-affinityECM binding state requires the actin cytoskeleton (6). In turn,integrin engagement with ECM induces signaling that mediatesactin polymerization and contractility downstream of Rho GTPases(6, 7). ECM-engaged integrins also affect cytoskeletal organi-zation by physically linking the contractile actomyosin system toextracellular anchorage points (7). Thus, adhesion-associatedactin structures are integrated systems that mediate cellular func-tions requiring coordination of intracellular cytoskeletal forces withECM binding.Mesenchymal cells generally possess two main types of adhe-

sion-associated actin structures: protruding lamellipodia con-taining nascent FAs at the cell edge and linear actin bundles inthe cell body connected to FAs. Compared with architecturally

invariant lamellipodia, adhesion-associated actin bundle struc-tures, including filopodia, the perinuclear actin cap/transmembraneactin-associated nuclear lines, trailing edge bundles, and dorsalSFs, are more diverse in their morphology and less well under-stood in their architecture and function (8–10). The most-studiedactin bundle structure is perhaps dorsal SFs, noncontractilebundles associated at one end with a ventral FA near the celledge and that extend radially toward the cell center and joinwith dorsal actin arcs on their other end. How the functionalspecificity of dorsal SFs is generated apart from the many otherdistinct adhesion-associated actin bundle structures is not wellunderstood.The functional specificity of adhesion-associated actin struc-

tures could be generated either on the adhesion side by com-positional differences in FA proteins or on the actin side bydifferences in the nucleation mechanism and actin binding pro-teins. On the adhesion side, it is well known that differentintegrin family members bind distinct types of ECM (11, 12).However, cells adhered to different ECMs all form commonstructures including lamellipodia, filopodia, and multiple typesof SFs. In addition to different integrins, FA function could beregulated by the process of “maturation” in which FAs undergostereotypical dynamic changes in composition and morphologydriven by actomyosin-mediated cellular tension (13, 14). NascentFAs contain integrins, focal adhesion kinase (FAK), a-actinin,

Significance

The ability of cells to interact with and remodel their extra-cellular environment is a critical process in developmentalmorphogenesis, wound healing, and cancer. How the physicaland chemical responses of fibroblasts to the extracellularmatrix are integrated across the cell remains a major openquestion. Previous data has shown a major role for the actincytoskeleton in coordinating deposition and organization ofthe extracellular matrix by fibroblasts. Our study examines therole of inverted formin 2 (INF2), a protein known to create newactin filaments, in mediating cellular response to extracellularconditions and control of extracellular matrix remodeling. Wefind that INF2 is responsible for generating specific actin struc-tures and specialized integrin-based fibrillar adhesions that arerequired for remodeling of the fibronectin extracellular matrixby fibroblasts.

Author contributions: C.T.S., A.D.D., and C.M.W. designed research; C.T.S. performed re-search; A.D.D. contributed new reagents/analytic tools; C.T.S. and S.V.P. analyzed data;and C.T.S. and C.M.W. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission. M.A.S. is a guest editor invited by the EditorialBoard.1To whom correspondence should be addressed. Email: watermancm@nhlbi.nih.gov.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1505035112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1505035112 PNAS | Published online April 27, 2015 | E2447–E2456

CELL

BIOLO

GY

PNASPL

US

See Retraction Published March 12, 2018

and paxillin (13, 15). When tension is applied, nascent FAs growand recruit hundreds of proteins, including talin, vinculin, andzyxin (16). These mature FAs then either disassemble or furthermature into tensin-containing fibrillar FAs that are responsiblefor fibronectin fibrillogenesis (17). Thus, the changes in FA sizeand protein content that accompany FA maturation could giverise to functional specialization of adhesion/actin systems.On the other hand, actin filaments in migrating cells are

generated by two main classes of nucleators: the Arp2/3 complexand formins (18). Different nucleating proteins generate differ-ent actin organization and geometries, which could in turndictate functional specificity of adhesions. Arp2/3 forms thebranched network in lamellipodia and is thought to be linked tonascent FAs through interaction with FAK (19–21) or vinculin(22). The formin family of actin nucleators, which generateslinear actin bundles (23), is more diverse, although formins sharea common actin assembly core domain (24), (25). Recent workhas begun to ascribe the generation of particular actin structuresto some of the 15 formins in mammalian cells, particularlymembers of the diaphanous family and FHOD1 (26–29). Spe-cifically regarding dorsal SFs, evidence points strongly to poly-merization by a formin family member (23, 30–32) but no forminhas ever been localized to these SFs or their associated FAs inmotile cells. Thus, although formins are clearly critical forforming distinct actin structures, whether they cooperate withFA proteins to specify the function of adhesion-associated actinstructures in the cell is unclear.We hypothesized that inverted formin 2 (INF2), found in our

recent FA proteome (33), may play a critical role in the forma-tion and functional specificity of adhesion-associated actinstructures. INF2 is expressed in cells in two isoforms, one con-taining a membrane-targeting CAAX-motif that plays a role inmitochondrial fission (34) and a non-CAAX isoform whosefunction is not well characterized. INF2 is an unusual formininsofar as it contains, in addition to the FH1–FH2 domains thatpolymerize actin, a WH2-like domain at the C terminus (35) thatbinds actin monomers to regulate autoinhibition, and also me-diates filament severing (35, 36). INF2 also interacts with andinhibits members of the diaphanous family of formin proteins(37). INF2 therefore could have multiple possible roles at FAs inlocal modulation of actin.Here we explore the role of INF2 in mouse embryonic fibro-

blasts (MEFs). We find for the first time to our knowledge stronglocalization of an endogenous formin to FAs at the distal tips ofdorsal SFs where it is required for actin polymerization at FAs toform dorsal SFs. We show that INF2 plays a role in controllingmorphological, but not compositional maturation of FAs.Strikingly, INF2 is responsible for the formation of one specificclass of FAs, the fibrillar FAs that organize the ECM; disruptionof INF2 leads to defects in ECM fibrillogenesis. Thus, our studydemonstrates that INF2 mediates the formation of dorsal SFsand fibrillar FAs, which together comprise a specific integratedadhesion-associated actin structure responsible for the fibrillo-genesis of ECM by fibroblasts.

ResultsThe INF2 Formin Localizes to Lamellipodia, Dorsal SFs, and FAs. Tounderstand the role of the formin INF2 in organization of theactin cytoskeleton in fibroblast function, we first sought to de-termine the actin structures with which it associates. PrimaryMEFs plated on fibronectin (FN)-coated coverslips were rapidlyfixed with acetone and costained for endogenous INF2, F-actin,and canonical markers of SFs (Fig. 1A) and subjected to spinningdisk confocal microscopy. As reported previously (34), INF2localized to narrow fibril structures in the center of the cellpresumed to be mitochondria (Fig. S1, blue arrow). Costainingof INF2 and F-actin showed that INF2 was restricted to a subset

of actin bundles in the lamella and to the lamellipodial actinmeshwork at the cell edge (Fig. 1 A–C).To determine the identity of the INF2-containing actin

bundles, we performed colocalization analysis with specific SFmarkers. Immunofluorescence analysis showed that INF2 local-ized along radial actin bundles that also contained α-actinin andtropomyosins (Fig. 1 A and E). These SFs extended perpendic-ularly to the leading edge and INF2 was concentrated at theirdistal tips where it colocalized with α-actinin. We then costainedfor myosin IIA, which is known to associate with transverse arcs

A B

C

D

Actin Tpm INF2 INF2/Tpm

Actin α-Actinin INF2 INF2/Atn

Bottom Top

Actin INF2

Actin MIIA INF2 INF2/MIIA

0.0

0.7

0.6

0.5

0.4

0.3

0.2

0.1

Actin α-Actinin Tpm

Stress Fiber ProteinsN.S.

N.S.*

Paxillin P-Tyr Vinculin0.0

0.7

0.6

0.5

0.4

0.3

0.2

0.1

Focal Adhesion Proteins

N.S.N.S

Paxillin INF2 Overlay

Vinculin INF2 Overlay

E

Actin INF2 Overlay

*

Fig. 1. Formin INF2 localizes to dorsal SFs at the junction with FAs.(A) Representative confocal micrographs of INF2 immunofluorescence (color-coded green) with Alexa-488 phalloidin staining of actin and immunofluo-rescence of the actin binding proteins α-actinin (Top, color-coded red), pan-tropomyosin (Middle, red) and myosin IIA (Bottom, red) in acetone-fixedMEFs. White boxes indicate area zoomed below. White arrowheads indicatecolocalization at actin structures. Red arrowheads indicate lack of colocali-zation. [Scale bar, (Top) 10 μm; (Bottom) 5 μm.] (B) Maximal projection ofconfocal Z series of Alexa-488 phalloidin staining of actin (Left) and immu-nofluorescence of INF2 (Right). Color scale of z position is shown at theLower Left. White arrowheads indicate colocalization along dorsal SFs. Ar-rows indicate dorsal actin structures (white) that lack INF2 (red). (Scale bar,10 μm.) (C) Representative confocal micrographs of Alexa-488 phalloidinstaining of actin (green) with INF2 (red) immunofluorescence with overlayshown. White boxes indicate area zoomed below. White arrowhead in-dicates colocalization of actin and INF2 at cell edge. [Scale bar, (Top) 10 μm;(Bottom) 5 μm.] (D) Representative confocal micrographs of INF2 immuno-fluorescence (green) and immunofluorescence of the focal FA proteins paxillin(Top, red) and vinculin (Bottom, red). White boxes indicate area zoomed be-low. White arrowheads indicate colocalization at FAs. [Scale bar, (Top) 10 μm;(Bottom) 5 μm.] (E) Bar graph of Pearson’s coefficient of colocalization be-tween INF2 and SF proteins (Top) or FA proteins (Bottom). n = 6 cells percondition. Error bars: SD. *P < 0.05; N.S., not significant, Student’s t test.

E2448 | www.pnas.org/cgi/doi/10.1073/pnas.1505035112 Skau et al.

See Retraction Published March 12, 2018

but is largely absent from dorsal SFs (23, 38). Myosin IIA lo-calized to arcs around the cell sides and actin bundles in the cellbody, but had reduced levels on INF2-decorated bundlesextending perpendicular to the edge (Fig. 1A, Bottom). Toconfirm that the SFs decorated with INF2 were in fact dorsalSFs, we generated 3D projections from confocal stacks andtracked these actin bundles in the Z direction from the ventralcell surface and found that they extended radially toward the cellcenter and up toward the dorsal cortex (Fig. 1B). EndogenousINF2 is therefore localized specifically to dorsal actin SFs inthe lamella.Because INF2 localization appeared to extend distally beyond

the termini of dorsal SFs, we hypothesized that INF2 was asso-ciated with FAs. To test our hypothesis, we examined colocali-zation of INF2 and FA proteins. INF2 partially colocalized withpaxillin at FAs in the lamellum, but extended more proximallythan the bulk of paxillin (Fig. 1 D and E). In contrast, INF2colocalized more strongly with the FA adaptor protein vinculin(Fig. 1 D and E), which localizes to the actin–FA interface (39).Together, these results show that in addition to mitochondria(34), endogenous INF2 localizes specifically to dorsal SFs andthe SF–FA junction in MEFs, as well as lamellipodia.

INF2 Is Required for Dorsal SFs in Lamellae. We next sought to de-termine the role of INF2 in organization of the actin cytoskele-ton. We used siRNA to decrease INF2 protein levels by 66%(Fig. 2A), then acetone-fixed and stained mock-transfectedcontrol and INF2 knockdown MEFs (INF2 KD, Fig. 2) withantibodies to INF2 and paxillin, and examined them by spin-ning disk confocal microscopy. These experiments showed thatsiRNA treatment resulted in loss of focal adhesion staining withthe anti-INF2 antibody in individual cells. Loss of INF2 consis-tently reduced cell area compared with controls (Fig. 2 B and C),possibly due to metabolic deficiencies induced by reduced INF2activity at mitochondria (34). Reexpression of the human INF2(and thus refractory to siRNA targeting the mouse protein)isoform lacking the mitochondrial targeting sequence fused at itsC terminus to EGFP (henceforth referred to as INF2-GFP) at121% of endogenous level showed that this isoform did not lo-calize to mitochondria in the cell center and failed to rescue thedefect in cell size, but localized to the cytoskeleton and FAsimilar to the endogenous protein (Fig. 2C and see Fig. 4), andrescued cytoskeletal and FA phenotypes induced by INF2 KD(see below). We therefore used this construct for the remainderof our studies.Despite the small size of INF2 KD cells, formaldehyde fixation

and staining with fluorescent phalloidin showed that they had awell-developed actin cytoskeleton, albeit with marked differ-ences in organization compared with mock-transfected controls.Control cells displayed a lamellipodium typified by a bright bandof dense actin at the cell edge and possessed dorsal and ventralSFs and transverse arcs in the lamellum (Fig. 2 B and D). Cellslacking INF2 possessed similar transverse arcs (Fig. 2B), butdisplayed a dramatic loss of actin bundles perpendicular to theleading edge in the lamella and cell center (Fig. 2 B–D), as wellas a slightly wider lamellipodium (Fig. S2).We focused our study on understanding the defect in SFs in

the absence of INF2. Compared with controls, INF2 KD cellslacked the SFs that normally extend radially through the lamellafrom the cell edge toward the top of the cell as seen in 3Dprojections, and also exhibited increased cell height (Fig. 2 B andC). Quantitative analysis showed that compared with control,INF2 KD significantly reduced the number and length of actinbundles (Fig. 2C). Reexpression of INF2-GFP in INF2 KDMEFs rescued SF number, but not their length, likely because oftheir smaller cell area (Fig. 2C). Remaining bundles in INF2 KDcells were localized along the sides and rear of the cell, althoughweak radial bundles were sometimes present in the lamella.

Immunostaining with the SF markers pan-tropomyosin, α-acti-nin, myosin II A, myosin II B, and phosphorylated myosin IIregulatory light chain showed that remaining actin bundles alongthe sides and rear of INF2 KD cells contained the appropriatecontractile machinery; however, the myosin II marker proteinswere largely absent from weak radial bundles, suggesting thatthey were remnants of dorsal SFs (Fig. 2D and Fig. S3). To-gether, these data indicate that INF2 is specifically required forformation of dorsal SFs to reduce cell height (40).

A

D Control MEFs INF2 KD MEFs

Actin Tm (pan) Overlay

α-ActininActin Overlay α-ActininActin Overlay

C

Avg

Siz

e, μ

m2

x 1

00

Cell Area

N.S.

****

Control INF2KD0

100

# S

F p

er

cell

2040

6080

# of Stress Fibers

Rescue Avg

Le

ng

th, μ

m

Stress Fiber Length

Actin Tm (pan) Overlay

N.S.**

** ****

**

Control MEFs INF2KD MEFs

Bottom TopActin Actin

B

Control MEF INF2 KD MEF

INF2 INF2Paxillin PaxillinOverlay Overlay

INF

2 P

rote

in, F

rac

Con h42 h84h51Con KD

1.6

0.0

1.20.80.4

INF2 siRNA

Resc

Bottom Top

35

05

1015202530

Control INF2KD Rescue Control INF2KD Rescue0

102030405060

Fig. 2. INF2 controls SF morphology and lamellipodial width. (A, Left Top)Bar graph of the average level of depletion and reexpression of INF2 in fourindependent experiments. Error bar: SD. (Bottom) Representative image ofWestern blot showing levels of INF2 in mock-transfected control (Con) andMEF transfected with siRNA targeting INF2 (INF2 KD) for 15, 24, and 48 h.(Right) Representative confocal micrographs of Alexa-488 phalloidin stain-ing of actin with immunofluorescence of INF2 in control and INF2 KD MEFshowing loss of INF2 at FA in INF2 KD. White boxes indicate area zoomedbelow. [Scale bar, (Top) 10 μm; (Bottom) 5 μm.] Note the difference in size ofthe scale bar in control and INF2 KD cells. (B) Black and white images:Representative confocal micrographs of Alexa-488 phalloidin staining ofactin in control and MEFs transfected with siRNAs targeting INF2 (INF2 KD).Color images: Maximal projection of confocal Z series of Alexa-488 phal-loidin staining of actin in control and INF2 KD MEFs. Color scale of z posi-tion is shown at Upper Right. (Scale bar, 10 μm.) (C) Bar graphs of theaverage cell area (Left), the number of SFs per cell (Center), and length ofSFs (Right) in control MEFs, INF2 KD MEFs, or INF2 KD MEFs reexpressing thehuman INF2 isoform lacking the mitochondrial targeting sequence fused toGFP (rescue) n = 20 cells per condition. Error bar: SD. **P < 0.01, N.S., notsignificant, Student’s t test. (D) Representative confocal micrographs ofAlexa-488 phalloidin (green) staining of actin with immunofluorescence ofthe actin-binding proteins α-actinin (Top, red) and pan-tropomyosin(Bottom, red) in control and INF2 KD MEFs. White boxes indicate areazoomed below. White arrowheads indicate linear actin bundles. [Scale bar,(Top) 10 μm; (Bottom) 5 μm.]

Skau et al. PNAS | Published online April 27, 2015 | E2449

CELL

BIOLO

GY

PNASPL

US

See Retraction Published March 12, 2018

INF2 Promotes Barbed End Formation and Actin Assembly at FA. Ourobservation that INF2 contributes to dorsal SF formation suggeststhat INF2 may regulate actin assembly specifically at FAs. To testthis prediction, we first determined the role of INF2 in generationof assembly-competent “free” barbed ends of actin filaments incontrol and INF2 KD cells (Fig. 3A). By gently permeabilizing cellsin the presence of rhodamine-labeled G-actin, fixing and costainingwith fluorescent phalloidin and an antibody against paxillin, wewere able to visualize sites of incorporation of new actin mono-mers onto free barbed ends, total F-actin, and FAs (Fig. 3A)(41). Spinning disk confocal images revealed incorporation ofrhodamine-actin indicating the location of free barbed ends alongthe leading edge of the lamellipodium and at FAs at the termini ofdorsal SFs in control cells (Fig. 3A). Quantification of the ratio ofrhodamine-actin to phalloidin fluorescence showed that comparedwith control, INF2 KD cells had a slight but insignificant reduc-tion in rhodamine-actin fluorescence in lamellipodia (Fig. S2E),but a strongly significant decrease in rhodamine-actin incor-poration at FAs (Fig. 3A). Therefore, INF2 promotes formation ofpolymerization-competent free barbed ends of actin filaments atFAs, whereas other mechanisms contribute to barbed end for-mation in lamellipodia.We then analyzed the role of INF2 in actin dynamics in dorsal

SFs at FAs in living cells. We coexpressed mApple-actin andEGFP-paxillin in MEFs in the presence and absence of siRNAstargeting INF2 and imaged cells by time-lapse TIRF microscopy

(Fig. 3B). Examination of movies of control cells showed pax-illin-containing nascent FAs formed within the lamellipodiumconcomitant with leading edge protrusion, as previously reported(Fig. 3B and Movies S1 and S2) (42). Nascent adhesions formingin the lamellipodium initially lacked actin bundles and the ma-jority underwent rapid disassembly within about a minute as thetrailing edge of the lamellipodium moved beyond them. For thenascent FAs that remained after the lamellipodium advanced, afine actin bundle appeared at the proximal end of the FAs andthe FAs began to elongate (Fig. 3B, Right and Movies S1 and S2).These linear bundles continued to extend from the proximal sideof the FAs as the FAs grew (Fig. 3B, Right). INF2 KDMEFs alsonucleated nascent FAs in protruding lamellipodia (Fig. 3B). Incontrast with control cells, however, these FAs failed to sproutactin bundles (Fig. 3B, Right and Movies S3 and S4). Rather thanlinear filaments, a dense actin meshwork associated with theseFAs (Fig. 3B). These structures remained associated with FAsfor many minutes after the edge advanced, but displayed little-to-no directional elongation (Fig. 3B, Right and Movies S3 and S4).Some actin aggregates in INF2 KD MEFs appeared to span orconnect multiple small FAs (Fig. 3B, Right), a phenomenon nottypical of dorsal SFs. The few weak radial actin structures weobserved were associated with the sides of a protrusion (Fig. 3B)or coalesced as multiple globules merged (Fig. 3B, Right). Theseresults show that INF2 is required for forming actin bundles atthe proximal edge of maturing FAs in the lamellum.

A

B

C Control (Actin-GFP)

INF2KD (Actin-GFP)

Control INF2KD0.0

0.2

0.4

0.6

0.8

Elo

ng

ati

on

Rat

e, μ

m/m

in

Bleach

BleachTime

Stress Fiber Elongation Rate

Control MEFs INF2KD MEFs Control MEFs

INF2KD MEFs

0 s 25 50 75 100 125 150 175 200 225 250

GFP-Pxn

mApple-Actin

Overlay

GFP-Pxn

mApple-Actin

Overlay

Control MEFs INF2KD MEFs

Paxillin

Actin

Barbed Ends

Overlay

Paxillin

Actin

Barbed Ends

Overlay

GFP-Pxn

mApple-Actin

Overlay

GFP-Pxn

mApple-Actin

Overlay

G-Actin Incorporation at FA

Control

No

rma

lize

d F

luo

resc

en

ce

0.0

0.4

0.1

0.2

0.3**

Rescue

**

**

Rescue (mApple-Actin) Bleach

N.S.

INF2 KD

-30 3000 30 60 90120150180210240270

-30 3000 30 60 90120150180210240270

-30 3000 30 60 90120150180210240270

Time

Time

0 s 25 50 75 100 125 150 175 200 225 250

Fig. 3. INF2 mediates barbed end formation and actin as-sembly at FAs. (A, Left) Representative confocal micrographs ofpermeabilized mock-transfected control and MEFs transfectedwith siRNAs targeting INF2 (INF2 KD) showing Alexa-488phalloidin (green) staining of F-actin together with X-rhoda-mine labeled G-actin (red) incorporation (barbed ends) andimmunofluorescence of paxillin (blue) with color overlayshown. White boxes indicate area zoomed below. White ar-rowheads indicate presence of G-actin incorporation at FAs.Red arrowhead indicates reduced incorporation. [Scale bar,(Left) 10 μm; (Right) 5 μm.] Note the difference in size of thescale bar in control and INF2 KD cells. (Right) Bar graph ofG-actin incorporation at FAs as measured by the ratio of X-rho-damine G-actin fluorescence to Alexa-488 phalloidin fluores-cence at FAs. n = 5 cells, at least 250 FAs per condition. Errorbar: SD. *P < 0.05; N.S., not significant, Student’s t test. (B, Left)Representative TIRF micrographs of control and INF2 KD MEFexpressing EGFP-paxillin (Top, green) and mApple-actin (Cen-ter, red). Color overlay is shown at Bottom. White boxes in-dicated area zoomed at Right. (Scale bar, 10 μm.) Right: TIRFtime lapse image sequences of an actin bundle (Center, red)elongating out of a paxillin-marked FA (Top, green) with coloroverlay shown (Bottom). Time in seconds is shown. White ar-rowheads indicate a single FA and associated actin. (Scale bar,2 μm.) (C, Left) Representative confocal time lapse image se-quence of EGFP-actin SF in mock-transfected or INF2 KD MEF,or mApple-actin SF in INF2 KD cells reexpressing the humanINF2-GFP isoform lacking the mitochondrial targeting se-quence (rescue). Red bar indicates area photobleached at time0. Time in seconds is shown. (Scale bar, 1 μm.) (Center) Kymo-graph projection of rearward movement/recovery of bleachedregion. Arrowhead indicates time of bleach. (Right) Bar graphof average SF elongation rate in mock transfected, INF2 KD, orINF2-GFP rescue. n = 10 SFs per condition. Error bar: SD. **P <0.01, Student’s t test.

E2450 | www.pnas.org/cgi/doi/10.1073/pnas.1505035112 Skau et al.

See Retraction Published March 12, 2018

To determine the role of INF2 in elongation of actin filamentsin dorsal SFs at FAs, we photobleached a narrow stripe across anEGFP-actin labeled SFs just proximal to its terminus at a FA,and monitored the bleached zone by time-lapse spinning diskconfocal microscopy (23). In control cells, the bleached area didnot recover fluorescence within 300 s, but during that timemoved away from the SF terminus at the FA at a consistent rate(Fig. 3C and Movie S5), indicating elongation of the actin bundledistal to the bleach mark at its site of attachment to the FA, aspreviously reported (23). Although INF2 KD induced loss ofmost dorsal SFs, we were able to bleach a stripe across remnantweak radial bundles attached to FAs in the lamella. This ex-periment showed that, like controls, the bleached mark in INF2KD cells did not recover fluorescence. However, in contrast tocontrol MEFs, in INF2 KD cells bleached marks on actin bun-dles remained almost stationary relative to the FA, moving awayfrom the FA at a significantly lower rate than control (Fig. 3Cand Movie S4). In INF2 KD cells reexpressing INF2-GFP,movement of the bleached SF stripe away from the FA was re-covered (Fig. 3C). Together, these data show that INF2 is re-quired for promoting actin polymerization and forming freebarbed ends at FAs to mediate formation of actin bundles atmaturing FAs and dorsal SFs.

INF2 Is Recruited to FAs During Myosin II-Dependent Maturation.Ourdata show that INF2 localizes at the FA–SF interface and pro-motes dorsal SF formation (Figs. 1–3). Because the SF interfacewith the FA forms as nascent FAs mature (42), we hypothesizedthat INF2 may assemble into FAs during maturation. To test thishypothesis, we determined the dynamic localization of INF2 toFAs during assembly and maturation. We cotransfected MEFswith INF2-GFP together with mApple- or mCherry-taggedpaxillin as a marker of FAs and performed time-lapse TIRFmicroscopy (Fig. 4 A–C). In control cells, mApple-paxillinappeared as diffraction-limited nascent FA punctae in pro-truding lamellipodia, most of which disassembled, but a fraction

of which exited the advancing lamellipodium and elongatedcentripetally as they matured (Fig. 4A and Movies S6 and S7).INF2-GFP localized to lamellipodia, making it difficult to di-rectly determine whether INF2 was at nascent FAs (Fig. 4A).However, fluorescence intensity line scans along the protrudingedge of cells revealed no specific coconcentration of INF2-GFPabove the lamellipodial localization in punctate mCherry-paxillin–labeled nascent FAs, and in regions just proximal tolamellipodia, young, slightly elongated FAs in the earliest stageof maturation contained little-to-no INF2-GFP (Fig. 4A, redarrowhead). Larger FAs located deeper in the lamella exhibitedhigh concentrations of INF2-GFP (Fig. 4 A and B). Two-colorkymograph analysis of individual FAs confirmed that nascentFAs exiting the advancing, INF2-rich lamellipodium lackedINF2-GFP (Fig. 4C, Inset and Movie S7), followed by strongINF2-GFP localization throughout the length of growing FAs,and finally specific concentration of INF2-GFP in the proximalend of fully mature FAs, presumably at the junction with SFs(Fig. 4C). Therefore, INF2 assembles into FAs as they mature.Proteins such as vinculin or zyxin that assemble into FAs

during maturation are recruited in a myosin II contractility-dependent manner (13, 43, 44). To determine if INF2 recruitmentto FAs was contractility dependent, we examined the response ofINF2 localization to inhibition of myosin II ATPase activity. Wetreated cells transfected with INF2-GFP with blebbistatin andstained them with antibodies to paxillin phosphorylated on Y31(phosphopaxillin) as a marker of immature FAs (45) and imagedthem by confocal microscopy. Blebbistatin treatment resultedin the formation of lamellipodia containing diffraction-limitedphosphopaxillin punctae and loss of mature FAs in the lamellaand cell center (Fig. 4D). Colocalization line scans showed thatin the absence of myosin II activity, INF2 was localized in arelatively uniform band along the cell edge but did not appearspecifically enriched in phosphopaxillin FA punctae (Fig. 4D).Together, these results show that INF2 is absent from nascent

B

INF2-GFP mApple-Pxn Overlay0 s 65 130195260325390455

INF2-GFP

mApple-Pxn

OverlayD

T

A

Phosphopaxillin INF2-GFP

Distance along edge, μm

Flu

ore

sce

nce

, a.u

.

0 15

INF2-GFP Phospho

paxillinOverlay

C

INF2-GFP mCherry-Pxn Overlay

0 5Dist. along edge

340

240

260

280

300

320

850

250

450

650

Flu

ore

sce

nce

, a.u

.

Nascent FA

Mature FA

Paxillin

Paxillin

INF2

INF2

D

Fig. 4. INF2 localizes to FAs during maturation. (A) Repre-sentative TIRF micrographs of MEFs cotransfected with humanINF2 isoform lacking the mitochondrial targeting sequencefused to EGFP (INF2-GFP, Left, green) and mCherry-paxillin(Center, red) with color overlay shown (Right). White boxesindicate area zoomed below. Arrowheads indicate a paxillin-marked FA (white) that lacks INF2-GFP (red). Blue and purplelines in the overlay indicate the position of line scans used forquantification of fluorescence intensities in nascent (Uppergraph in B) and mature (Lower graph in B) FAs, respectively.[Scale bar, (Top) 3 μm; (Bottom) 2 μm.] (B) Fluorescence in-tensities of mCherry-paxillin (red) and INF2-GFP (green) alongthe lines in A. (C, Left) Representative TIRF micrographs ofMEFs cotransfected with INF2-GFP (Left, green) and mApple-paxillin (Center, red) with color overlay shown. White box in-dicates the area shown in the kymograph in the Inset as well asin the zoomed time-lapse series at Right. (Inset) Kymographsperpendicular to the leading edge across the FA are shown inthe white box. D, distance; T, time. White arrowheads indicateearliest appearance of INF2-GFP. (Scale bar, 10 μm.) (Right)Time-lapse TIRF time lapse images of the cell edge and FAmarked by INF2-GFP (green, Top) and paxillin (red, Center).Time in seconds is indicated. (Scale bar, 1 μm.) White arrowheadindicates an FA born at the cell edge. (D) Representative confocalmicrographs of MEF transfected with INF2-GFP (Left, green)treated for 2 h with 50 μm blebbistatin then fixed and stainedfor paxillin phosphorylated on tyrosine 31 (phosphopaxillin, red,Center) with color overlay shown. White box indicates areazoomed below. White arrowhead indicates a paxillin-marked FAthat lacks INF2-GFP (red arrowhead). Blue line in Left indicatesthe position of the line scan used for quantification of fluores-cence intensities of phospho-paxillin and INF2-GFP in nascent FAon the graph at Right. [Scale bar, (Top) 10 μm; (Bottom) 2 μm.]

Skau et al. PNAS | Published online April 27, 2015 | E2451

CELL

BIOLO

GY

PNASPL

US

See Retraction Published March 12, 2018

FAs, but is recruited to FAs during myosin II-dependent FAgrowth and maturation.

INF2 Promotes Formation of Large and Fibrillar FAs.Our observationthat INF2 is recruited during FA growth suggests that it may berequired for FA maturation. To test this hypothesis, we exam-ined the role of INF2 in FA morphology, composition, and dy-namics (Figs. 5 and 6). First, to characterize FA morphology, westained cells in the presence (INF2 KD) or absence (control) ofsiRNAs targeting INF2 for actin and paxillin. Compared withcontrol, INF2 KD cells lacked extended FAs in the cell center,and their peripheral FAs were smaller (Fig. 5 A–D). Rather thanexisting as a very narrow band at the cell edge, these small FAswere present further into the lamellum than in control cells (Fig.5 A and E). Analysis of FA area confirmed this finding, showingthat compared with control, INF2 KD cells had significantlysmaller FAs (Fig. 5 A and B). Furthermore, in controls, approxi-mately one-third of the total FAs were nascent FAs (<0.3 μm2) andtwo-thirds were medium (0.3–2.0 μm2) or large matureFAs (>2.0 μm2, Fig. 5C), whereas INF2 KD shifted the distri-bution such that over 80% of the FAs were nascent and greatlydecreased the fraction of medium and large FAs. In control cells,FAs were oblong in shape (aspect ratio > 3), whereas FAs inINF2 KD were significantly rounder and less extended (Fig. 5D).FA size was rescued by expression of INF2-GFP in INF2 KDcells, although reexpression did not fully rescue FA aspect ratio,likely because the cells remained smaller than average (Figs. 2 Aand E, 5 B and D, and 6G). Thus, INF2 reduces nascent FAs andpromotes large, oblong FAs in the cell center.Because we found that INF2 increases FA size, we hypothe-

sized that INF2 may also promote the protein compositionalchanges that accompany FA growth during maturation. Wecostained control and INF2 KD cells for paxillin and markers ofnascent (FAK, refs. 13, 15), mature (vinculin and zyxin, refs. 13,15), and fibrillar (tensin, refs. 15, 17) FAs and imaged them byconfocal microscopy (Fig. 5E and Fig. S4A). Control cells pos-sessed diffraction-limited nascent FAs at their edges markedby paxillin, vasodilator-stimulated phosphoprotein (VASP), andFAK, whereas small oblong FAs at the lamellipodium–lamellumborder as well as extended FAs in the lamellum also containedvinculin and zyxin (Fig. 5E and Fig. S4A). Staining for tensinshowed strong staining of long FAs in the cell center and dim orvariable staining of small FAs in the cell periphery (Fig. 5E). InINF2 KD cells, the small round FAs in the cell peripheryexhibited the same complement of proteins seen in nascent andsmall FAs of control cells, including paxillin, FAK, vinculin,zyxin, and variable levels of tensin (Fig. 5E and Fig. S4A).However, INF2 KD cells completely lacked tensin-containingextended FAs in the cell center (Fig. 5E). Localization of tensinto oblong FAs formed in INF2 KD cells reexpressing INF2-GFPwas recovered, although due to small cell size, fewer FAs wereobserved in the cell center (Fig. 5F). Thus, INF2 is not requiredfor compositional maturation of FAs, but is critical for forminglarge, central, tensin-containing FAs.To determine if the signaling properties of FAs are modulated

by INF2, we examined tyrosine phosphorylation of FA proteinsthat occurs in response to integrin engagement to ECM. Wecoimmunolocalized paxillin together with either total phospho-tyrosine or with antibodies specific to FAK phosphorylated onY397 (phospho-FAK) or phosphopaxillin in control and INF2KD cells (Fig. S4B). Confocal imaging showed that FAs in bothcontrol and INF2 KD cells contained high levels of tyrosinephosphorylation, phospho-FAK, and phosphopaxillin (Fig. S4B).Together, these results show that INF2 reduces the fraction ofnascent FAs and is also required for formation of fibrillar FAs inthe cell center, but not recruitment of FA proteins or basic as-pects of integrin signaling.

Control MEFs INF2KD MEFs

Control MEFs

Pxn Vinc Overlay

Pxn Tensin Overlay

INF2KD MEFs

<0.3 um2 0.3-

2.0 um2

INF2 KD0.8

0.6

0.4

0.2

0.0Fra

c o

f A

dh

esi

on

s

0.0

2.0

1.5

1.0

0.5

Avg Adhesion Size

Ad

he

sio

n S

ize

, um

2

0Con INF2KD

Avg Aspect Ratio

Asp

ect

Rat

io

1

2

3

4

5

Rescue

Adhesion Distribution

** ***

** **N.S.

Con INF2KD Rescue

INF2-GFP Rescue MEFs

FAK

Tensin INF/TensINF2-GFPActin

Control

>2.0 um2

Actin Paxillin Overlay Actin Paxillin Overlay

VASP Overlay

FAK VASP Overlay

Pxn Vinc Overlay

Pxn Tensin Overlay

A

B C D

E

F

Fig. 5. INF2 is critical for morphological maturation of FAs. (A) Representativeconfocal micrographs of mock-transfected control MEF and MEF transfected withsiRNAs targeting INF2 (INF2 KD) showing Alexa-488 phalloidin staining of actin(green) with immunofluorescence of paxillin (red). White boxes indicate areazoomed below.White arrowheads show termination of SF at FA. [Scale bar, (Top)10μm;(Bottom)5μm.]Notethedifferenceinsizeofthescalebar incontrolandINF2KD cells. (B–D) Quantification of FA morphometry from analysis of fluorescenceimages of paxillin in control, INF2 KD, and INF2 KD cells reexpressing the humanINF2-GFP isoform lacking the mitochondrial targeting sequence (rescue). n = 15cells, at least 900 FAs per condition. (B andD) Bar graphs of FA area (B) and aspectratio (D). Error bars: SD. **P < 0.01; *P < 0.05; N.S., not significant, Student’s t test.(C) Box and whisker plot of FA size distribution, means are indicated by red bars.(E)Representativeconfocalmicrographsof immunofluorescenceofpaxillin (Pxn)orVASP (both green) with FA proteins found at nascent (FAK, Top, red), maturing[vinculin (Vinc),Middle, red] and fibrillar (tensin, Bottom, red) FAs in control (Left)and INF2KD(Right)MEFs.Whiteboxes indicateareaszoomedbelow;whitedashedboxes indicate area zoomed on bottom row. White arrows indicate nascent FA,arrowheads denote mature FA, red arrow indicates lack of colocalization oftensin with paxillin in nascent FA. [Scale bar, (Top) 10 μm; (Bottom Left) 5 μm.](F) Immunofluorescence of tensin and INF2-GFP in INF2-GFP re-expressingMEF.White boxes indicate areas zoomed below. Arrowheads denote colocalizationof tensin and INF2 at central adhesion. [Scale bar, (Top) 10 μm; (Bottom) 5 μm.]

E2452 | www.pnas.org/cgi/doi/10.1073/pnas.1505035112 Skau et al.

See Retraction Published March 12, 2018

INF2 Promotes Turnover of Nascent FAs and Is Critical to FA Growthand Elongation. To determine how INF2 inhibits nascent FAs andpromotes fibrillar FAs, we performed time-lapse TIRF micros-copy of FA dynamics and quantitative image analysis (Fig. 6 A–F).First, we tracked the overall dynamic behavior of FAs bycreating color-encoded time overlays that compared GFP-paxillin–marked FAs near the cell edge before and after a 120-s timeinterval (Fig. 6A). In this analysis, purple represents FAs newlyassembled during the 120-s interval, green shows FAs that dis-assembled during the time interval, and white FAs were constantthroughout (Fig. 6A and Movie S8). This showed that in 120 s incontrol MEFs, many FAs formed near the leading edge andmany FAs disassembled in the lamella, but few remained con-stant, suggesting a rapid FA assembly and turnover cycle. InINF2 KD MEFs, although many FAs formed near the leadingedge in 120 s, most of them remained constant and very fewturned over in the lamella, indicating a reduction in FA turn-over compared with control. Coexpression of INF2-GFP andmCherry paxillin in INF2 KD cells rescued rapid FA assemblyand turnover dynamics (Fig. 6B and Movie S8). Time-lapse im-age series at 5-s intervals and kymograph analysis showed that incontrol MEFs, most nascent FAs turned over rapidly as theleading edge and lamellipodium advanced, whereas a subsetunderwent rapid centripetal elongation as they matured in thelamellum (Fig. 6 C and D and Movie S9) (42). In contrast, mostof the nascent FAs in INF2 KD MEFs remained round punctaeven as the leading edge advanced and new FAs were nucleatedin front of them. The few FAs that did grow in INF2 KD cellseither expanded radially as two neighboring punctate FAsmerged or grew centripetally much more slowly than control cells(Fig. 6D and Movie S9) and never extended into the cell center.To determine the precise defect in FA dynamics, we performedquantitative analysis of FA nucleation rate (number of nascentFAs formed/unit time/unit lamellipodial area), maturation frac-

tion (the fraction of nascent FAs that do not disassemble in theleading edge but go on to elongate), and lifetime (time fromappearance to disappearance of individual FAs) on image seriesacquired at 5-s intervals. This analysis showed that, as suspectedfrom examination of time overlays, there was no difference in FAnucleation rate in control and INF2 KD MEFs (Fig. 6E). Oncenascent FAs nucleated, nearly 25% of them matured in thelamellum of control cells, whereas in INF2 KD cells only 5% ofnascent FAs matured in this fashion (Fig. 6F). Furthermore, thelack of nascent FA turnover in INF2 KD cells resulted in asignificant threefold increase in FA lifetime compared withcontrol (Fig. 6G). Together these results suggest that INF2 re-duces nascent FAs not by inhibiting their formation, but bypromoting both their turnover and transition to rapid elongationand likely their eventual transition to central fibrillar FAs.

INF2 Regulates Traction Force and ECM Fibrillogenesis. Our resultsshow that INF2 is critical to regulation of the actin cytoskeletonand FAs. To determine the physiological significance of thesefunctions of INF2, we examined the requirement for INF2 intraction force and generation of ECM by fibroblasts. First,we used high-resolution traction-force microscopy (TFM) (46)to determine the force exerted by control and INF2 KDcells expressing GFP-paxillin by plating on compliant, 8.6-kPaFN-coupled polyacrylamide gels embedded with fluorescent beadsto serve as fiducial marks of cell-generated substrate displacement(Fig. 7A). FAs in control and INF2 KD cells were plated oncompliant substrates resembled FAs in cells on glass; FAs inINF2 KD cells were smaller than those in control cells (Fig. 7A).TFM analysis showed that cells lacking INF2 generated signifi-cantly higher stress on the substrate than controls (Fig. 7A). Thisdemonstrates that INF2 attenuates traction force on the ECM,in agreement with the notion that large FAs exert little force onthe substrate compared with nascent FAs (47).

A

INF

2 K

D

0 s

0 s

120 s

120 s

BINF2KD+INF2-GFP

INF2-GFP; mCherry-Pxn

mCherry-Paxillin

INF2-GFP

0 s

0 s

150 s

150 s

0 s/150s

0 s/150s

0 s/120 s

0 s/120 s

0 s 50 100 150 200 250 300

0 s 50 100 150 200 250 300

TD

TD

Co

ntr

ol

INF

2 K

D

C D

FEMaturation Fraction

0.25

0.00

0.200.150.100.05

Fra

c th

at

Elo

ng

ate

GAdhesion Lifetime

0

LIfe

tim

e, m

in

123456

**

Control

Nucleation Rate1.0

NA

nu

cle

ate

d/

10

0 s

ec/

μm

2

N.S.0.350.30 **

Control ControlINF2 KD INF2 KD INF2 KD0.0

0.2

0.4

0.6

0.8

GFP-Pxn

GFP-Pxn

Co

ntr

ol

Fig. 6. INF2 promotes centripetal elongation andturnover of nascent FAs. (A) TIRF micrographs fromtime-lapse image series of the cell edge in mock-transfected control and MEF transfected with siR-NAs targeting INF2 (INF2 KD) expressing EGFP-paxillin. The color overlay shows the two timepoints 120 s apart; purple FAs were assembled in thetime gap, green FAs disassembled during the timegap, and white FAs remained present throughoutthe 120-time gap. (Scale bar, 2 μm.) (B, Left) Rep-resentative TIRF micrograph of INF2 KD cells reex-pressing the human INF2-GFP isoform lacking themitochondrial targeting sequence (INF2-GFP, green)and mCherry-paxillin (red). (Right) TIRF micrographsfrom time-lapse image series of the cell edge incontrol and INF2 KD MEFs expressing INF2-GFP.The overlay shows the two time points differen-tially color encoded [t = 0 s (green) and t = 150 s(purple)]. (Scale bar, 5 μm.) (C ) RepresentativeTIRF micrographs of control and INF2 KD MEFtransfected with EGFP-paxillin. White rectanglehighlights the region shown in time-lapse se-quence in D. Dashed white box highlights the re-gion shown in overlay in A. (Scale bar, 10 μm.)(D) TIRF time-lapse micrographs of FA dynamics incontrol and INF2 KD MEF; FA marker EGFP-paxillin;time in seconds is shown. Dashed red line indicateslines along FA used for kymograph analyses (Right)of FA growth. D, distance; T, time. (Scale bar, 1 μm.)Dashed white line highlights the slope, indicativeof the rate of FA growth. (E–G) Bar graphs of na-scent FA formation rate per micrometer of cell edge(nucleation rate, E), the fraction of nascent FAs thatundergo maturation (maturation fraction, F), and average FA lifetime (G), in control and INF2 KD MEF. FA marker, paxillin. n = 5 cells, at least 200 ad-hesions per condition. Error bar: SD. **P < 0.01; N.S., not significant, Student’s t test.

Skau et al. PNAS | Published online April 27, 2015 | E2453

CELL

BIOLO

GY

PNASPL

US

See Retraction Published March 12, 2018

The major physiological function of fibroblasts, deposition andorganization of FN-containing ECM, is mediated by tensin-positive fibrillar FAs (48). The reduced FA size and unusualdistribution of tensin in cells lacking INF2 suggests INF2 may becritical for ECM fibrillogenesis. We first tested the requirementfor INF2 in the ability of human foreskin fibroblasts (HFFs) togenerate cell-derived matrices (CDMs). We used HFFs as op-posed to MEFs because shRNA plasmids to human INF2 con-tinually inhibited INF2 expression for the long time periodsrequired for CDM generation, compared with the transientknockdown obtained for siRNA in MEFs. Cells expressing eitherthe GFP-tagged shRNA or GFP alone as a control were platedat confluency for 7 d and supplemented with ascorbic acid topromote deposition of ECM, after which cells were lysed andwashed out, leaving behind CDMs (Fig. 7B). Phase-contrastimaging showed that the CDM deposited by control cells washomogeneous and dense, with thick, phase-dense bundles ofaligned and interconnected fibrils (Fig. 7B). Conversely, com-pared with control, the CDM produced by cells lacking INF2 wasmore heterogeneous in density and isotropic in organization,with regions completely lacking thick fibrils and much lessinterconnected mesh (Fig. 7B).We then determined the requirement for INF2 in FN fibril-

logenesis. We plated control and INF2 KD cells on glass,allowed them to deposit and reorganize FN for 24 h, then fixedand costained for FN and paxillin to determine the organizationof FN relative to FAs (Fig. 7C). Control cells deposited FN andorganized it into fibrils associated with paxillin-containing FAstoward the center of the cell (Fig. 7C). Cells lacking INF2,however, were unable to form FN fibrils. Some FN accumulatedaround INF2 KD cells, but it remained diffuse in the cell center

or formed irregular blobs in the periphery that were not stronglyassociated with FAs (Fig. 7C). To determine if the defect ofINF2 KD cells in fibrillogenesis was due to lack of FN secretionor an inability to fibrillarize plasma FN, we assayed the ability ofliving cells to incorporate fluorescently labeled FN added to themedia into ECM fibrils. This showed that control cells efficientlyassembled fluorescent FN into fibrillar structures close to largecentral FAs at the termini of SFs (Fig. 7D, Top row). In contrast,INF2 knockdown cells exhibited a lack of fluorescent FN fibrilsunder the cell, although hazy fluorescence was present (Fig. 7D,Middle row). Reexpression of INF2-GFP in INF2 KD cells res-cued the defect, with these cells regaining the ability to form FNfibrils at the termini of SFs in the center of the cell (Fig. 7D,Bottom row). Together, these data show that INF2 is required forthe ability of fibroblasts to organize ECM and fibrillarize FN onboth long and short timescales.

DiscussionOur results show for the first time to our knowledge that theformin INF2 localizes specifically to FAs, dorsal SFs, andlamellipodia in fibroblasts. Although it is well accepted thatformin proteins act at FAs to mediate SF formation (23, 30–32),the ability to localize a formin to FAs has eluded the field. Thislocalization is quite distinct from the CAAX membrane-targetedisoform of INF2, which localizes to mitochondria and the en-doplasmic reticulum (34, 49). We find that FA-associated INF2plays a critical role in promoting formation of polymerization-competent free actin barbed ends that mediate the centripetalelongation of an FA-associated actin bundle to form a dorsal SF.INF2 assembles into FAs in a contractility-dependent mannerduring FA maturation rather than during initial FA generation,

A

B C Control MEFs

Actin Paxillin FibronectinOverlay

INF2KD MEFs

Control CDM INF2KD CDM

D

Control INF2 KD Rescue0

908070605040302010Fl

uo

resc

en

ce a

.u.

**N.S.

Co

ntr

ol

INF

2 K

DR

esc

ue

Actin FN Overlay

Control INF2 KD0.0

1.0

0.8

0.6

0.4

0.2

No

rma

lize

d T

ract

ion

Fo

rce

,

nN

/μm

2

Stress Exerted by Cells

*

Traction Force Microscopy

INF2KDControlGFP-Paxillin

Extracellular Matrix Production

Actin Paxillin Fibronectin Overlay

Actin FN Overlay

Actin FN Overlay

Fig. 7. INF2 regulates traction force and ECM fibrillogenesis.(A, Left) Representative confocal images of EGFP-paxillin inmock-transfected control and MEF transfected with siRNAstargeting INF2 (INF2 KD) plated on fiducial-marked poly-acrylamide substrates of stiffness, 8.6 kPa. (Scale bar, 10 μm.)Note the difference in size of the scale bar in control and INF2KD cells. (Right) Bar graph of normalized traction force gen-erated on substrates by control and INF2 KD cells as de-termined by Fourier transform traction microscopy (46). *P <0.05, Student’s t test. n = 5 cells per condition. (B) Represen-tative phase-contrast images of cell-derived matrix producedby mock-transfected control human foreskin fibroblasts (HFFs)or HFFs transfected with shRNA targeting INF2. White boxesindicate area zoomed below. [Scale bar, (Top) 400 μm; (Bot-tom) 100 μm.] (C) Representative confocal micrographs ofAlexa-488 phalloidin staining of actin (green) with immuno-fluorescence of paxillin (red) and FN (blue) in control and INF2KD MEF. White boxes indicate area zoomed below. White ar-rowheads indicate fibrils of FN near FA, red arrowhead in-dicates diffuse FN. [Scale bar, (Top) 20 μm; (Bottom) 5 μm.]Note the difference in size of the scale bar in control and INF2KD cells. (D, Left) Representative confocal micrographs ofcontrol MEF, INF2 KD MEF, or INF2-KD MEF reexpressing INF2-GFP (rescue) transfected with mApple-actin (green) and platedon FN-coated coverslips. Images were acquired 6 h after Alexa-647–labeled FN (red) was added to the media and allowed tobe incorporated into fibrils. White boxes indicate area zoomedbelow. White arrowheads indicate accumulation of labeled FNat the termini of stress fibers; red arrowhead indicates lack ofFN accumulation. [Scale bar, (Left) 10 μm; (Right) 5 μm.] (Right)Graph of average labeled FN fluorescence under each cellnormalized to cell area. n = 10 cells per condition, error bar =SD. *P < 0.05; N.S., not significant, Student’s t test.

E2454 | www.pnas.org/cgi/doi/10.1073/pnas.1505035112 Skau et al.

See Retraction Published March 12, 2018

and once there, stimulates the formation of elongated FAs andis absolutely required for formation of large, central, tensin-containing fibrillar FAs. We show that the major role of INF2is in endowing fibroblasts with the ability to organize FN intoECM on both long and short timescales. Collectively our re-sults indicate an important role for the formin INF2 in spec-ifying the function of fibrillar FAs through its ability togenerate dorsal SFs. Thus, dorsal SFs and fibrillar FAs form aspecific class of integrated adhesion-associated actin struc-ture in fibroblasts that mediates their critical physiologicalrole in ECM generation.How is INF2 promoting dorsal SF formation and FA matu-

ration? Because previous studies have shown that nascent FAmaturation requires a bundled actin template (42) and FAs incells lacking INF2 lack actin bundles, this suggests that the roleof INF2 in FA maturation is through regulation of FA-associatedactin rather than direct effects on FAs. INF2 has a number ofactivities that could modulate actin at FAs, including the FH1–FH2 domains that polymerize actin, a WH2-like domain at the Cterminus (35) that binds actin monomers to regulate auto-inhibition, and also mediates filament severing (35, 36). INF2also interacts with and inhibits members of the diaphanousfamily of formins (37). The simplest scenario would be that theactin polymerization activity of INF2 mediates barbed end for-mation and actin filament assembly at FAs, which would act asan actin bundle template for FA maturation (42) and drive SFassembly. However, a transient association with Dia1 and/orDia2 could “prime” INF2 for localization at FAs and explain thelongstanding conundrum of the fact that the diaphanous forminsare required for actin assembly at FAs, yet are not localized tothese sites (23, 30–32, 37, 50). Similarly, we cannot rule out a rolefor the severing activity of INF2 (35, 51) in increasing barbedends at FAs as well as throughout dorsal SFs, a phenomenonalso not fully understood (23, 41). Finally, we also consider thefact that INF2 generates stable detyrosinated microtubules thatcan help drive cell migration through centrosome positioning(52). Dissecting the mechanism by which INF2 mediates SFformation and FA maturation will require structure–functionanalysis to determine the FA targeting domain and generation ofactivity-specific mutants for reconstitution in cells lacking INF2.Our data also weigh in on the controversy regarding the roles of

actin assembly versus myosin II contractility in FA maturation. Wefind that INF2 is required for barbed end formation and directionalfilament elongation at FAs and also required for FA maturation(42). This finding supports the notion that FA maturation requiresan assembling actin bundle template, and that actomyosin tensionthroughout the network is not sufficient for full FA maturation(Figs. 2 and 3). Supporting this finding, recent studies have high-lighted the role of actin polymerization, rather than myosin II ac-tivity, in FA maturation (32, 53). As tension on formins has beenshown to increase their associated actin filament assembly rate,an alternative role for myosin-mediated tension could be toenhance polymerization of actin SFs at FAs (32, 53–55). Inaddition, our results demonstrating that knockdown of INF2decreases the FA maturation fraction without affecting thenascent FA nucleation rate further highlights the critical roleof actin assembly in regulating the decision point betweennascent FA turnover and maturation, suggesting that assemblyof the FA-associated actin bundle stabilizes nascent FA againstdisassembly and promotes the transition to FA growth.Our examination of the composition of FAs in the absence of

dorsal SFs allowed us to uncover the surprising finding thatmorphological and compositional maturation of FAs are notdirectly coupled, as has been assumed by the field (56). Ca-nonical models of FA maturation couple elongation/growth ofnascent FAs to changes in protein phosphorylation and accu-mulation of additional proteins (for review see ref. 14). We findthat in the absence of dorsal SFs generated by INF2, FAs largely

do not elongate into oblong FAs, resulting in a lack of large FAsin the center of the cell as has been seen in other cells lackingdorsal SFs (Figs. 5 and 6) (32, 42, 57). Surprisingly, however,these morphologically immature FAs at the cell edge containproteins generally associated with mature FAs, such as zyxin andvinculin and a variable amount of tensin, and are competent forintegrin signaling. Therefore, we suggest that a major role of SFsis to couple morphological and compositional maturation ofFAs. Our findings also call to question the role of FA compo-sitional maturation in general, and more specifically the role ofthe protein tensin, in generating functional fibrillar FAs. We suggestthat extended morphology of oblong FAs controls the ability of FAsto assemble ECM, and not the presence of tensin per se. Here, weshow that dorsal SFs direct this morphological and functionalswitch, likely by promoting directional elongation of FAs.We find that in addition to FAs, INF2 also localizes to the

protruding edge of migrating cells just proximal to the actin-nucleating Arp2/3 complex. However, unlike Arp2/3, whichcreates the dendritic lamellipodial actin network, we find thatINF2 limits lamellipodial width and rate of actin retrograde flowthrough unknown mechanisms. This finding is consistent withdata showing that INF2 can inhibit actin assembly by diaphanousfamily formins (37), which have been proposed to play a role inprotrusion of lamellipodia (31, 32, 58), and this interaction mayalso provide a mechanism for recruitment of INF2 to the leadingedge. However, based on recent evidence showing competitionbetween actin nucleation factors in cells (59), we also hypothe-size that INF2 may limit lamellipodial width by competing withother actin nucleation factors for monomers at the cell edge. Inaddition, although INF2 is not localized specifically in nascentFAs in lamellipodia, we cannot rule out an indirect role oflamellipodial INF2 in affecting nascent FA dynamics.On a cellular scale, the physiological role of SFs generated by

INF2 also remains incompletely understood. Some data indicatethat prominent SFs actually inhibit rapid cell motility (32, 60).Indeed, we do not see major defects in cell migration even whendorsal SFs are significantly reduced by loss of INF2 (Fig. S5). Wesuggest that by ensuring proper FA maturation, dorsal SFs couldalso contribute to proper force balance in cells; when only smallhigh-tension FAs at the cell edge are present, cells exhibit milddefects in persistent spreading and migration. However, theimportance of SFs in generating tension or turning off tension atFAs remains unclear. Our data suggest that a critical role for thedorsal SF-adhesion structure is in fibrillarizing FN mediated bythe formin INF2. Our data thus not only reveal the previouslyunidentified ability of an actin nucleation factor to specifyphysiological functions of primary fibroblasts, but also expandthe role for the SF–FA interaction.

Materials and MethodsMEFs were isolated and maintained as described previously (54). Forcontrol and INF2 KD fixation, cells were fixed with paraformaldehyde asdescribed previously (61). For endogenous INF2 staining, cells were fixedwith ice-cold acetone at −20° for 10 min then washed with TBS, thenblocking and staining were performed as previously described (61). De-tection of free barbed ends was performed as described previously (41)incubating 0.5 μM rhodamine-actin on cells for 2 min 15 s. Chemicals andantibodies were obtained from commercial sources (SI Materials andMethods) except for antitensin antibody (a gift from Benjamin Geiger,Weizmann Institute of Science, Rehovot, Israel) and antizyxin antibody (agift from Mary Beckerle, Huntsman Cancer Institute, Salt Lake City). Imagingwas performed as described previously (54) with the addition of a MYO cooledCCD camera (Photometrics) for some confocal imaging. Image analysis wasperformed by hand except for Pearson’s coefficient determination by JACoP(62). TFM was performed as described previously (46, 63). Cell-derived matriceswere generated as described previously (64). See SI Materials and Methods foradditional details.

ACKNOWLEDGMENTS. The authors thank Dr. Mike Davidson and Dr. HenryHiggs for fluorescent protein constructs; Dr. Benjamin Geiger for the

Skau et al. PNAS | Published online April 27, 2015 | E2455

CELL

BIOLO

GY

PNASPL

US

See Retraction Published March 12, 2018

antitensin antibody; Dr. Mary Beckerle for the antizyxin antibody; Dr. IngoThievessen and Dr. Robert Fischer for mouse pedagogy and helpfuldiscussion; William Shin for work on the C.M.W. laboratory microscopes;

and Schwanna Thacker for administrative assistance. This work is sup-ported by the Intramural Program of the National Heart, Lung, and BloodInstitute, NIH.

1. Wu C, et al. (2013) Loss of Arp2/3 induces an NF-κB-dependent, nonautonomous ef-fect on chemotactic signaling. J Cell Biol 203(6):907–916.

2. Case LB, Waterman CM (2011) Adhesive F-actin waves: A novel integrin-mediatedadhesion complex coupled to ventral actin polymerization. PLoS ONE 6(11):e26631.

3. Martinelli R, et al. (2013) Release of cellular tension signals self-restorative ventrallamellipodia to heal barrier micro-wounds. J Cell Biol 201(3):449–465.

4. Lewellyn L, Cetera M, Horne-Badovinac S (2013) Misshapen decreases integrin levelsto promote epithelial motility and planar polarity in Drosophila. J Cell Biol 200(6):721–729.

5. Rozario T, DeSimone DW (2010) The extracellular matrix in development and mor-phogenesis: A dynamic view. Dev Biol 341(1):126–140.

6. Takagi J, Springer TA (2002) Integrin activation and structural rearrangement. Im-munol Rev 186:141–163.

7. Machesky LM, Hall A (1997) Role of actin polymerization and adhesion to extracel-lular matrix in Rac- and Rho-induced cytoskeletal reorganization. J Cell Biol 138(4):913–926.

8. Galbraith CG, Yamada KM, Galbraith JA (2007) Polymerizing actin fibers positionintegrins primed to probe for adhesion sites. Science 315(5814):992–995.

9. Kim D-H, et al. (2012) Actin cap associated focal adhesions and their distinct role incellular mechanosensing. Sci Rep 2:555.

10. Zand MS, Albrecht-Buehler G (1989) What structures, besides adhesions, preventspread cells from rounding up? Cell Motil Cytoskeleton 13(3):195–211.

11. Byron A, Morgan MR, Humphries MJ (2010) Adhesion signalling complexes. Curr Biol20(24):R1063–R1067.

12. Schwartz MA (2001) Integrin signaling revisited. Trends Cell Biol 11(12):466–470.13. Zaidel-Bar R, Ballestrem C, Kam Z, Geiger B (2003) Early molecular events in the as-

sembly of matrix adhesions at the leading edge of migrating cells. J Cell Sci 116(Pt 22):4605–4613.

14. Gardel ML, Schneider IC, Aratyn-Schaus Y, Waterman CM (2010) Mechanical in-tegration of actin and adhesion dynamics in cell migration. Annu Rev Cell Dev Biol 26:315–333.

15. Zamir E, et al. (1999) Molecular diversity of cell-matrix adhesions. J Cell Sci 112(Pt 11):1655–1669.

16. Zaidel-Bar R, Cohen M, Addadi L, Geiger B (2004) Hierarchical assembly of cell-matrixadhesion complexes. Biochem Soc Trans 32(Pt3):416–420.

17. Katz BZ, et al. (2000) Physical state of the extracellular matrix regulates thestructure and molecular composition of cell-matrix adhesions. Mol Biol Cell 11(3):1047–1060.

18. Waterman CM, Skau C (2015) Specification of architecture and function of actinstructures by actin nucleation factors. Ann Rev Biophys, in press.

19. Le Clainche C, Carlier M-F (2008) Regulation of actin assembly associated with pro-trusion and adhesion in cell migration. Physiol Rev 88(2):489–513.

20. Koestler SA, et al. (2013) Arp2/3 complex is essential for actin network treadmillingas well as for targeting of capping protein and cofilin. Mol Biol Cell 24(18):2861–2875.

21. Serrels B, et al. (2007) Focal adhesion kinase controls actin assembly via a FERM-mediated interaction with the Arp2/3 complex. Nat Cell Biol 9(9):1046–1056.

22. DeMali KA, Barlow CA, Burridge K (2002) Recruitment of the Arp2/3 complex tovinculin: Coupling membrane protrusion to matrix adhesion. J Cell Biol 159(5):881–891.

23. Hotulainen P, Lappalainen P (2006) Stress fibers are generated by two distinct actinassembly mechanisms in motile cells. J Cell Biol 173(3):383–394.

24. Krainer EC, et al. (2013) The multiplicity of human formins: Expression patterns in cellsand tissues. Cytoskeleton (Hoboken) 70(8):424–438.

25. Schönichen A, Geyer M (2010) Fifteen formins for an actin filament: A molecular viewon the regulation of human formins. Biochim Biophys Acta 1803(2):152–163.

26. Svitkina TM, et al. (2003) Mechanism of filopodia initiation by reorganization of adendritic network. J Cell Biol 160(3):409–421.

27. Schirenbeck A, Arasada R, Bretschneider T, Schleicher M, Faix J (2005) Formins andVASPs may co-operate in the formation of filopodia. Biochem Soc Trans 33(Pt 6):1256–1259.

28. Kutscheidt S, et al. (2014) FHOD1 interaction with nesprin-2G mediates TAN lineformation and nuclear movement. Nat Cell Biol 16(7):708–715.

29. Schulze N, et al. (2014) FHOD1 regulates stress fiber organization by controlling thedynamics of transverse arcs and dorsal fibers. J Cell Sci 127(Pt 7):1379–1393.

30. Riveline D, et al. (2001) Focal contacts as mechanosensors: Externally applied localmechanical force induces growth of focal contacts by an mDia1-dependent andROCK-independent mechanism. J Cell Biol 153(6):1175–1186.

31. Gupton SL, Eisenmann K, Alberts AS, Waterman-Storer CM (2007) mDia2 regulatesactin and focal adhesion dynamics and organization in the lamella for efficient epi-thelial cell migration. J Cell Sci 120(Pt 19):3475–3487.

32. Oakes PW, Beckham Y, Stricker J, Gardel ML (2012) Tension is required but not suf-ficient for focal adhesion maturation without a stress fiber template. J Cell Biol196(3):363–374.

33. Kuo J-C, Han X, Hsiao C-T, Yates JR, 3rd, Waterman CM (2011) Analysis of the myosin-II-responsive focal adhesion proteome reveals a role for β-Pix in negative regulationof focal adhesion maturation. Nat Cell Biol 13(4):383–393.

34. Korobova F, Ramabhadran V, Higgs HN (2013) An actin-dependent step in mito-chondrial fission mediated by the ER-associated formin INF2. Science 339(6118):464–467.

35. Chhabra ES, Higgs HN (2006) INF2 Is a WASP homology 2 motif-containing forminthat severs actin filaments and accelerates both polymerization and depolymerization.J Biol Chem 281(36):26754–26767.

36. Ramabhadran V, Hatch AL, Higgs HN (2013) Actin monomers activate inverted formin2 by competing with its autoinhibitory interaction. J Biol Chem 288(37):26847–26855.

37. Sun H, Schlondorff JS, Brown EJ, Higgs HN, Pollak MR (2011) Rho activation of mDiaformins is modulated by an interaction with inverted formin 2 (INF2). Proc Natl AcadSci USA 108(7):2933–2938.

38. Tojkander S, et al. (2011) A molecular pathway for myosin II recruitment to stressfibers. Curr Biol 21(7):539–550.

39. Kanchanawong P, et al. (2010) Nanoscale architecture of integrin-based cell adhe-sions. Nature 468(7323):580–584.

40. Burnette DT, et al. (2014) A contractile and counterbalancing adhesion system con-trols the 3D shape of crawling cells. J Cell Biol 205(1):83–96.

41. Symons MH, Mitchison TJ (1991) Control of actin polymerization in live and per-meabilized fibroblasts. J Cell Biol 114(3):503–513.

42. Choi CK, et al. (2008) Actin and alpha-actinin orchestrate the assembly and matura-tion of nascent adhesions in a myosin II motor-independent manner. Nat Cell Biol10(9):1039–1050.

43. Hirata H, Tatsumi H, Sokabe M (2008) Mechanical forces facilitate actin polymeriza-tion at focal adhesions in a zyxin-dependent manner. J Cell Sci 121(Pt 17):2795–2804.

44. Dumbauld DW, et al. (2010) Contractility modulates cell adhesion strengtheningthrough focal adhesion kinase and assembly of vinculin-containing focal adhesions.J Cell Physiol 223(3):746–756.

45. Nayal A, et al. (2006) Paxillin phosphorylation at Ser273 localizes a GIT1-PIX-PAKcomplex and regulates adhesion and protrusion dynamics. J Cell Biol 173(4):587–589.

46. Sabass B, Gardel ML, Waterman CM, Schwarz US (2008) High resolution traction forcemicroscopy based on experimental and computational advances. Biophys J 94(1):207–220.

47. Beningo KA, Dembo M, Kaverina I, Small JV, Wang YL (2001) Nascent focal adhesionsare responsible for the generation of strong propulsive forces in migrating fibro-blasts. J Cell Biol 153(4):881–888.

48. Sottile J, Hocking DC (2002) Fibronectin polymerization regulates the compositionand stability of extracellular matrix fibrils and cell-matrix adhesions. Mol Biol Cell13(10):3546–3559.

49. Chhabra ES, Ramabhadran V, Gerber SA, Higgs HN (2009) INF2 is an endoplasmicreticulum-associated formin protein. J Cell Sci 122(Pt 9):1430–1440.

50. Sun H, Schlondorff J, Higgs HN, Pollak MR (2013) Inverted formin 2 regulates actindynamics by antagonizing Rho/diaphanous-related formin signaling. J Am SocNephrol 24(6):917–929.

51. Gurel PS, et al. (2014) INF2-mediated severing through actin filament encirclementand disruption. Curr Biol 24(2):156–164.

52. Andrés-Delgado L, et al. (2012) INF2 promotes the formation of detyrosinated mi-crotubules necessary for centrosome reorientation in T cells. J Cell Biol 198(6):1025–1037.

53. Stricker J, Beckham Y, Davidson MW, Gardel ML (2013) Myosin II-mediated focaladhesion maturation is tension insensitive. PLoS ONE 8(7):e70652.

54. Thievessen I, et al. (2013) Vinculin-actin interaction couples actin retrograde flow tofocal adhesions, but is dispensable for focal adhesion growth. J Cell Biol 202(1):163–177.

55. Feng Y, et al. (2013) α-Actinin1 and 4 tyrosine phosphorylation is critical for stressfiber establishment, maintenance and focal adhesion maturation. Exp Cell Res 319(8):1124–1135.

56. Balaban NQ, et al. (2001) Force and focal adhesion assembly: A close relationshipstudied using elastic micropatterned substrates. Nat Cell Biol 3(5):466–472.

57. Bershadsky AD, et al. (2006) Assembly and mechanosensory function of focal adhe-sions: Experiments and models. Eur J Cell Biol 85(3-4):165–173.

58. Kurokawa K, Matsuda M (2005) Localized RhoA activation as a requirement for theinduction of membrane ruffling. Mol Biol Cell 16(9):4294–4303.

59. Burke TA, et al. (2014) Homeostatic actin cytoskeleton networks are regulated byassembly factor competition for monomers. Curr Biol 24(5):579–585.

60. Couchman JR, Rees DA (1979) The behaviour of fibroblasts migrating from chick heartexplants: Changes in adhesion, locomotion and growth, and in the distribution ofactomyosin and fibronectin. J Cell Sci 39:149–165.

61. Pasapera AM, Schneider IC, Rericha E, Schlaepfer DD, Waterman CM (2010) Myosin IIactivity regulates vinculin recruitment to focal adhesions through FAK-mediatedpaxillin phosphorylation. J Cell Biol 188(6):877–890.

62. Bolte S, Cordelières FP (2006) A guided tour into subcellular colocalization analysis inlight microscopy. J Microsc 224(Pt 3):213–232.

63. Gardel ML, et al. (2008) Traction stress in focal adhesions correlates biphasically withactin retrograde flow speed. J Cell Biol 183(6):999–1005.

64. Kutys ML, Doyle AD, Yamada KM (2013) Regulation of cell adhesion and migration bycell-derived matrices. Exp Cell Res 319(16):2434–2439.

E2456 | www.pnas.org/cgi/doi/10.1073/pnas.1505035112 Skau et al.

See Retraction Published March 12, 2018