* CorrespondE-mail addr

1760-2734/$ -doi:10.1016/j.

Journal of Veterinary Cardiology (2012) 14, 211e221

www.elsevier.com/locate/jvc

Interactions between TGFb1 and cyclic strain inmodulation of myofibroblastic differentiation ofcanine mitral valve interstitial cells in 3D culture

Andrew S. Waxman, DVM a, Bruce G. Kornreich, DVM, PhD a,Russell A. Gould, BS, MS b, N. Sydney Moı̈se, DVM, MS a,Jonathan T. Butcher, PhD b,*

aClinical Sciences, College of Veterinary Medicine, Cornell University, Ithaca, NY 14853, USAbBiomedical Engineering, Cornell University, Ithaca, NY 14853, USA

KEYWORDSFibroblast;Strain;Myxomatous degenera-tion;Tissue engineering;Dog

ing author.ess: Jtb47@cornell.edu

see front matter ª 201jvc.2012.02.006

Abstract Objectives: The mechanisms of myxomatous valve degeneration (MVD)are poorly understood. Transforming growth factor-beta1 (TGFb1) induces myofi-broblastic activation in mitral valve interstitial cells (MVIC) in static 2D culture,but the roles of more physiological 3D matrix and cyclic mechanical strain areunclear. In this paper, we test the hypothesis that cyclic strain and TGFb1 interactto modify MVIC phenotype in 3D culture.Animals, materials and methods: MVIC were isolated from dogs with and withoutMVD and cultured for 7 days in type 1 collagen hydrogels with and without 5 ng/ml TGFb1. MVIC with MVD were subjected to 15% cyclic equibiaxial strain with staticcultures serving as controls. Myofibroblastic phenotype was assessed via 3D matrixcompaction, cell morphology, and expression of myofibroblastic (TGFb3, alpha-smooth muscle actin e aSMA) and fibroblastic (vimentin) markers.Results: Exogenous TGFb1 increased matrix compaction by canine MVIC with andwithout MVD, which correlated with increased cell spreading and elongation. TGFb1increased aSMA and TGFb3 gene expression, but not vimentin expression, in 15%cyclically stretched MVIC. Conversely, 15% cyclic strain significantly increased vi-mentin protein and gene expression, but not aSMA or TGFb3. 15% cyclic strainhowever was unable to counteract the effects of TGFb1 stimulation on MVIC.Conclusions: These results suggest that TGFb1 induces myofibroblastic differentia-tion (MVD phenotype) of canine MVIC in 3D culture, while 15% cyclic strain promotesa more fibroblastic phenotype. Mechanical and biochemical interactions likely regu-late MVIC phenotype with dose dependence. 3D culture systems can systematicallyinvestigate these phenomena and identify their underlying molecular mechanisms.ª 2012 Elsevier B.V. All rights reserved.

(J.T. Butcher).

2 Elsevier B.V. All rights reser

ved.

Abbreviations

3D 3-dimensionalMVD myxomatous valve degenerationMVIC mitral valve interstitial cellsaSMA alpha-smooth muscle actinTGFb1-3 transforming growth factor-beta1-3

c Worthington Biochemical Corp., Lakewood, NJ.d Invitrogen Corp., Grand Island, NY.e Hycel Inc., Houston, TX.

212 A.S. Waxman et al.

Introduction

Mechanical stress and biochemical alterations cancontribute to the phenotypic change of the mitralvalve undergoing degeneration.1,2 Understandinghow these factors contribute to myxomatous valvedegeneration (MVD) in the dog, demands explora-tion not only in vivo, but also through in vitro studiesthat permit specific examination of mechanical andsignaling mechanisms. Mitral valve leaflets arehighly organized, layered structures populated byinterstitial cells (MVIC) that are responsible fordeveloping and maintaining tissue stability andmatrix architecture.3,4 This stability and architec-ture is markedly destroyed in the myxomatousleaflets5 Normally MVIC are fibroblastic in cyto-skeletal phenotype, but promote a high degree ofmatrix turnover to support critically importantvalve function within their extremely demandinghemodynamic environment.6 In both canine andhuman valves with MVD, MVIC transition to a myofi-broblastic-like cell expressing contractile filamentssuch as alpha-smooth muscle actin (aSMA) anddesmin.7e9 The destruction of organized collage-nous matrix in mitral valves with MVD, combinedwith deposition of glycosaminoglycans, results insignificantlymore compliant leaflet tissues,which ispostulated to contribute to their insufficiency.10 It iswell known that mitral valve leaflets are subjectedto multi-axial cyclic deformation during the cardiaccycle,11,12 but how tissue mechanics contribute tocanine MVIC phenotype and valve matrix remodel-ing is less clear.

Cytokines such as transforming growth factorbeta (TGFb) have been implicated in the patho-genesis of canine MVD.13,14 Elevated TGFb1-3 geneand protein expression have been found in caninemitral valves with MVD.13,14 Components of sero-tonin metabolism are also differentially expressedin canine MVD,15 which has been shown to modu-late TGFb signaling and myofibroblastic activationin sheep and porcine valve interstitial cells.16

Recent studies have shown that alterations incyclic mechanical strain can lead to MVIC trans-formation to a myofibroblast-like, that is, MVDphenotype.17,18 Cyclic stretch also alters

glycosaminoglycan and proteoglycan profiles inporcine MVIC seeded collagen gels.19e21 The amountof collagen and glycosaminoglycan synthesis hasbeen shown to be dependent on the amplitude andfrequencyof stretch.19e21 It has been suspected thatextracellular matrix remodeling takes place inresponse to localmechanical stimuli and ismediatedbyTGFb1.22,23 Many of the strain inducedeffects canbe reversed by TGFb1 signaling blockade.26 Merry-man et al17 demonstrated in porcine aortic valveinterstitial cells that cyclic strain and exogenousTGFb synergize to elevate collagen synthesis andTGFb signaling. These findings suggest that cyclicmechanical strain and TGFb signaling may acttogether to promote mitral valve remodeling, butwhether and how this occurs in the context of canineMVD is unknown. In this study, we test the hypoth-eses that TGFb1 and cyclic mechanical strainmodulatemyofibroblastic differentiation andmatrixcompaction by canine MVIC in 3-dimensional (3D)culture. We further probe the synergistic or antag-onistic effects of strain and TGFb on canine MVIC.

Methods

Tissue collection and cell culture

Anterior mitral valve leaflets were collected frombeagle dogs euthanized for reasons as part ofinstitutionally approved in vitro studies unrelatedto mitral valves. The dogs were healthy and hadbeen given no medications. Specimens were clas-sified as having either normal (n ¼ 2) or mildlymyxomatous (n ¼ 2) degeneration via gross path-ological inspection with the latter characterizedby thickened, irregular leaflets and mitral regur-gitation documented by echocardiography. Bothdogs with MVD were 7 years of age while thenormal dogs were 2 years of age. Cells from thenormal dogs were used in the compaction studiesalone. For all other studies only the cells from themyxomatous valves were used. Mitral valve inter-stitial cells were isolated from the non-chordalregions of each sample independently via colla-genase digestion (Type II, 300 U/ml)c as previouslydescribed.18 Cells were grown on tissue culturetreated polystyrene and fed with Dulbecco’smodified Eagle’s medium (DMEM)d with 10% fetalbovine serume and 1% penicillin/streptomycin.d

Media was exchanged every 48 h, and cells werepassaged upon confluence. Cells were used inassays between passages 3 and 5.

TGFb1 and cyclic strain in myofibroblastic differentiation 213

Matrix compaction

Trypsinized confluent MVIC (normal MIVIC ormyxomatous MVIC) were pelleted via centrifuga-tion and dispersed within neutralized type 1collagen hydrogelsf (2 mg/ml) at a concentrationof 1 � 106 cells/ml as previously described.24 Thesolution was inoculated into culture well plates,creating cylindrical gels of uniform initial area.After 1 h of solidification, gels were released fromtheir culture substrate and fed either culturemedia alone or media supplemented with 5 ng/mlof human recombinant TGFb1.g Gels were thenallowed to culture free-floating for 7 days, withdigital images taken daily. From these images,hydrogel cross-section area was calculated usingImageJ.h Matrix compaction was then expressed asa ratio of final to original area.

Equibiaxial strain of MVIC hydrogels

Mitral valve leaflets in vivo experience a complexheterogeneous cyclic biaxial strain profile25 that isdifficult to impose in vitro. We approximated thisstrain environment using a novel cyclic strainbioreactor capable of stimulating engineered tissuemodels.26 (Fig. 1) Briefly as we have described morethoroughly elsewhere,26 cylindrical wells werefabricated within thick elastomeric silicone (1:10ratio of catalyst:PDMS) slabs. These slabs were thenaffixed between two aluminum plates, leavingconcentric circular openings below each well. Theplates were then adhered to a stage whose verticalmotion was controlled by a screw gear and rotarystepper motor. As the stage is raised and lowered,each well is stretched across a platen, creatinga homogeneous equiaxial strain distribution witha sinusoidal time profile. The entire system ishoused in a standard tissue culture incubator, whichis maintained at 37 �C and 5% CO2 throughout theculture period. For these experiments,1 � 106 cells/ml were suspended in 3D collagenconstructs as detailed above and 150 ml of solutionwas placed in each well. Each silicone elastomerwell is lined by a stainless steel spring for stable butflexible adhesion of the hydrogel. After 1 h of gelsolidification, each well was supplemented withculture medium. Constructs were allowed tocompact for an additional 24 h without strain.Medium was changed at 24 h and replaced witheither standard medium or medium supplementedwith 5 ng/ml of TGFb1. For all strain experiments,

f BecktoneDickinson, Franklin Lakes, NJ.g R&D Systems, Inc., Minneapolis, MN.h ImageJ NIH, Bethesda, MD.

15% area strain (equibiaxial) at a frequency of 1 Hzwas applied for 48 h, with static cultures serving ascontrols.

Cell phenotype assessment

At the completion of experiments, myxomatousMVIC constructs were fixed in 4% poly-formaldehyde. Cell morphology in response tocyclic strain in 3D culture was determined viaphalloidin based f-actin filament staining aspreviously described,28 using cell area and circu-larity as metrics.11 Immunofluorescent antibodystaining for aSMA and vimentin protein expressionindicated myofibroblastic or fibroblastic pheno-types as previously described.27 Briefly, constructswere washed with phosphate buffered saline, thenpermeabilized with 0.2% Triton-X 100 for 10 min.Another wash with phosphate buffered saline wasfollowed by overnight blocking with 1% bovineserum albumin kept at 4 �C. The blocking solutionwas aspirated and 1:100 dilutions of rabbit anti-vimentin and mouse anti-aSMA were added andincubated overnight at 4 �C. After additionalwashes, secondary antibodies (goat anti-mouse568 and goat anti-rabbit 488) at 1:100 dilutionswere added to constructs and incubated for 2 h atroom temperature. Draq5 was used at 1:1000dilution as a DNA counterstain. Fluorescence wasvisualized utilizing confocal microscopy (Leica,LSM510).i Maximal projections of a stacked seriesof images (10 � 10 um) were created using Leicasoftware.i The total number of cells was calcu-lated by converting images to greyscale. Thepercentages of cells positive for vimentin andaSMA were obtained for each condition usingImageJ,h with Draq5 positive nuclei as an indicatorof total cells. Analyze particles were used toselect each nucleus and count the total number ofcells. Ratios were made for the number of cellspositive for a particular protein to the totalnumber of cells. Data was normalized to staticcontrols.

Real-time PCR

Real-time PCR was conducted to assess geneexpression changes in myxomatous MVIC withcyclic strain and/or TGFb1. Primers for genesencoding proteins associated with myofibroblastphenotype (ACTA2, TGFb3), fibroblast phenotype(vimentin), and a housekeeping gene (GAPDH)were designed using Primer3 (MIT) and validatedusing canine testicular cDNA (Table 1). After the

i Leica Microsystems, Wetzlar, Germany.

Figure 1 System for cyclic equibiaxial strain of 3D cultured cells. (A) Translation stage containing 4 cylindricalcassettes (e.g. top left and bottom right) each with 4 holes concentric above circular platens (B). The cassettessandwich a thick slab of elastomeric material, within which small partial depth wells are created (C) into whicha hydrogel is cast. The hydrogel containing cells is anchored by means of a spring steel ring that flexes as the well isstretched, thus stretching the gel (D). In D the inner jagged ring symbolizes the nonstretched hydrogel and the outerjagged ring the stretched. The central schematic demonstrates the mechanism of cyclic stage translation and gelstretch. See Gould et al, Acta Biomaterialia 2012, for more details.

214 A.S. Waxman et al.

conclusion of experiments, gels were lysed in RLTbuffer and mRNA isolated using the RNEasy kitj

according to the manufacturer’s instructions andstored at �80 �C until use. cDNA was then createdfrom mRNA using First Strand RT-PCR kit.d Real-time PCR amplification was achieved using theSYBR-green systemk and read on an Optimaxthermocycler.k The Livak methodk (DDCT method)was used to calculate the fold change compared toGAPDH gene controls. Gene expression data wasthen expressed relative to unstrained gels culturedwithout TGFb1. Confidence intervals for the pointestimate of the fold change were calculated aspreviously described.l

j Qiagen Inc., Valencia, CA.k Biorad Laboratories, Hercules, CA.l http://www3.appliedbiosystems.com/cms/groups/mcb_

support/documents/generaldocuments/cms_042380 AppliedBiosystems, 2008. Guide to performing relative quantitation ofgene expression using real-time quantitative PCR. Part No.4371095 Rev B.

Statistical analysis

3D compaction was compared between the normaland myxomatous MVIC, while the effects of strain,TGFb1, and strainþTGFb1 on ACTA2, TGFb3 orvimentin expression was tested only in mildlymyxomatous MVIC. Data are presented for theseanalyses with medians and ranges (median:range)and with dot-plots as applicable. For each of themultiple comparisons the Kruskal Wallis Test wasused to determine if a difference existed betweengroups. The multiple comparisons included: (1)compaction between normal and myxomatousMVIC with and without TGFb3, (2) area, circularityindex, and (3) expression levels of ACTA2, TGFb3,and vimentin under control, strain, TGFb1, andstrain/TGFb1. If the Kruskal Wallis Test detecteda significant difference in the multiple compari-sons or for two nonpaired comparisons, analysisusing a ManneWhitney U test was performed witha Bonferroni correction for multiple comparisons.m

m Analytical Software, Tallahassee FL.

Table 1 Primers for canine genes encoding proteins associated with myofibroblast phenotype (ACTA2, TGFb3),fibroblast phenoptype (vimentin), and housekeeping gene (GAPDH) used in this study.

Gene Primer sequence Product Length (bp)

GAPDH 50-TGGCAAAGTGGATATTGTCG-30 14930-AGATGGACTTCCCGTTGATG-50

Vimentin 50-TGGCAAAGTGGATATTGTCG-30 15030-AGATGGACTTCCCGTTGATG-50

ACTA2 50-CCCAGACATCAGGGAGTGAT-30 14130-CTTTTCCATGTCGTCCCAGT-50

TGFb3 50-CTTGCACCACCTTGGACTTT-30 14830-CTGTTGTAAAGGGCCAGGAC-50

TGFb1 and cyclic strain in myofibroblastic differentiation 215

For each of the 3 genes, we set the comparison-wise a at 0.05. Publically available software wasutilized for analysis (PAST).n The gene expressionsare presented as fold differences þ/�95% confi-dence intervals.

Results

Hydrogel compaction of normal and myxo-matous canine MVIC

Canine MVIC compacted collagen hydrogels overthe 7-day period, with final gel area 20e40% oforiginal area depending on the experimentalcondition. As seen in Fig. 2 the compaction (day 7area/initial area) for the diseased MVIC hydrogelschanges rapidly between days 2 and 3. At experi-mental endpoint of day 7 (Fig. 2), myxomatousMVIC in the presence of TGFb1 (n ¼ 8) (0.179;0.161e0.212) compacted hydrogels more thanmyxomatous MVIC without TGFb1 (n ¼ 7) (0.112;0.109e0.118) (P ¼ 0.005). The compaction bynormal MVIC with TGFb1 (n ¼ 2) (0.323;319e0.326) did not differ from that of normal MVICwithout TGFb1 (n ¼ 5) (0.185; 0.179e0.221) withmultiple pairwise comparisons, but the lownumbers likely affected this analysis.

Effects of TGFb1 and strain on myxomatousMVIC cell morphology

A total of 30 cells were measured per gel, with fourhydrogels per treatment group, and 4 treatmentgroups were tested (120 cells per treatment). Cellsize/structure was quantified via area and thecircularity index.20 For the circularity indexa value of 0 represented a line and a value of 1

n http://folk.uio.no/ohammer/past. Hammer O, Harper DATand Ryan PD, 2001. PAST: Paleontological Statistics softwarepackage for education and data analysis. Palaeontologia Elec-tronica 4(1):9.

represented a perfect circle. Collectively, the areaof MVIC changed (P ¼ 0.005) with TGFb1 (290.54;230.83e313.03 pixels), strain (63.02; 42.26e82.76pixels), and TGFb1 plus strain (155.65;88.50e192.90 pixels) compared to that of MVICwith no TGFb1 and no strain (67.04; 43.43e97.00pixels), but post hoc pairwise comparisons couldnot identify a statistical difference among treat-ment groups. (Fig. 3A) Moreover, the circularityindex of MVIC changed (P ¼ 0.02) with TGFb1(0.250; 0.238e0.283), strain (0.429; 0.356e0.622),and TGFb1 plus strain (0.269; 0.240e0.288)compared to that of MVIC with no TGFb1 and nostrain (0.609; 0.499e0.665), but post hoc pairwisecomparisons could not identify a statisticaldifference among treatment groups (Fig. 3B). Thisis likely due to the small number of hydrogels ineach group (n ¼ 4). Fig. 3C and 3D illustrate whythe median MVIC area is greater and the medianMVIC circularity index is lower for MVIC cellstreated with TGFb1 than without TGFb1. That is,without TGFb1 the MVIC in 3D culture did not havean elongated morphology, but were more polyg-onal with a random orientation (Fig. 3C). Whentreated with 5 ng/ml TGFb1, MVIC appear largerand elongated, with filipodia like processesextending from the polar ends of the cells witha random orientation (Fig. 3D).

Effects of TGFb1 and strain on myxomatousMVIC protein and gene expression

In static 3D culture, MVIC expressed both vimentinand aSMA (Fig. 4). Unfortunately, an insufficientnumber of TGFb1 treated samples stained properlyfor protein expression quantification. However, wedid find that as a result of strain, vimentin proteinexpression significantly increased (P ¼ 0.026)(0.837; 0.758e0.871 versus 0.658; 0.512e0.728% innon-strained), while aSMA protein expression wasunchanged (P ¼ 0.68) (0.556; 0.535e0.631 versus0; 0.438e0.594% in non-strained) (Fig. 3B and C).Collectively, these findings suggest that mildly

Figure 2 Compaction of 3D hydrogels by mitral valve interstitial cells (MVIC) from normal valves and valves withmild myxomatous degeneration. Gels of each cell source were cultured with or without TGFb1. Compaction wasquantified as projected area with respect to original (day 0) area. A significant compaction amongst the groups wasnoted at day 7 with the pairwise comparisons revealing a significant difference only between the treated and non-treated myxomatous cell hydrogels.

216 A.S. Waxman et al.

myxomatous MVIC are somewhat myofibroblast-like in static 3D culture, but that 15% cyclic equi-biaxial strain promotes fibroblast-likedifferentiation.

Compared with static controls (n ¼ 7) alter-ations in gene expression of ACTA2 (gene forprotein aSMA), TGFb3, and vimentin wereobserved when cyclic strain (n ¼ 5), TGFb1 (n ¼ 8),or both (n ¼ 8) were applied to 3D cultures. (Fig. 5,Table 2) ACTA2 expression did not differ fromcontrol in TGFb1 or strain plus TGFb1 treated 3Dcultures; however, TGFb1 increased ACTA2expression when compared to cells subjected tocyclic strain alone (P ¼ 0.04). The expression ofTGFb3 with only cyclic strain applied was insig-nificantly less than control, but the expression ofTGFb3 was significantly increased with exogenousTGFb1 administration both with (P ¼ 0.009) andwithout (P ¼ 0.02) cyclic strain. Strain did not alter

TGFb3 expression and the application of strain didnot inhibit the expression of more TGFb3 whentreated with TGFb1. Expression of vimentin wassignificantly enhanced by strain alone (P ¼ 0.03),but not by TGFb1 alone. In fact, when 3D cultureswere treated with both TGFb1 and cyclic strain,the expression of vimentin was significantlyreduced compared to that observed in 3D culturestreated with strain alone.

Discussion

Our findings suggest that 5 ng/ml TGFb1 promotestransdifferentiation of canine MVIC to a myofibro-blast phenotype in 3D cultures as assayed byfunctional collagen gel compaction, cellmorphology, and gene expression of aSMA, andTGFb3. The functional effects as assessed by

Figure 3 TGFb1 alters morphology of myxomatous MVIC in 3D culture. Dot-plots of area (A) and circularity index (B)show changes with 15% cyclic strain and TGFb1. (C,D): Immunofluorescent images of 3D culture cell morphology (via f-actin) without (C) and with (D) TGFb1.

TGFb1 and cyclic strain in myofibroblastic differentiation 217

compaction studies of TGFb1 on canine MVICphenotype appear similar in normal canine MVICand MVIC obtained from dogs with MVD, althoughthese effects appear to be more profound inmyxomatous MVIC. Matrix compaction is drivenprimarily by cell-mediated traction forces.28 Ourresults indicate that normal MVIC treated withTGFb1 exert similar traction forces than myxoma-tous MVIC without TGFb1. In addition, we haveshown that 15% cyclic equibiaxial strain inducesa transformation of canine MVIC to a fibroblastphenotype in 3D cultures, as assayed by proteinexpression of vimentin and aSMA and cellmorphology. Finally, our results suggest thatalthough cyclic strain may temper the morphologicresponse induced by TGFb on canine MVIC, theeffect of TGFb1 on canine MVIC phenotype domi-nate that of cyclic strain. These findings validate

our experimental model as a means of investi-gating the biochemical and mechanical influenceson canine MVIC phenotype and provide insight intothe interaction of strain and TGFb1 on canine MVICin 3D culture.

Canine MVD is a complex disease that is similarin many respects to MVD in humans.29 Discoveringthe molecular and physical mechanisms thatpromote the transition from normal valves topathological mitral valve disease in dogs is there-fore an important research effort. The use of 3Dmatrix culture systems is important in recapitu-lating perhaps more physiological cellematrixinteractions that are vital to valve cell func-tion.24,30 In this study, we employ 3D collagenhydrogels to create 3D culture models of mitralvalve tissues using canine MVIC. Matrix compactionis a function of active cell contraction31,32 and cell

Figure 4 Protein expression changes in canine MVIC with cyclic stretch and/or TGFb1. (A) Immunofluorescentstaining for alpha-smooth muscle actin (aSMA, red), vimentin (green), and DNA (blue) in MVIC cultured in 3Dhydrogels. Cells were from 3D cultures that were under static control conditions (0,0), treatment with TGFb1 alone(0,þ), 15% cyclic strain alone (þ,0) or both TGFb1 and strain (þ,þ). Cells expressing aSMA (B) and vimentin (C) werequantified and normalized to total cell number.

218 A.S. Waxman et al.

traction forces associated with migration.33 Ourresults suggest that myxomatous MVIC haveenhanced migratory and/or traction force gener-ation capacity. A number of studies suggest thatcontractile protein expression increases in MVIC

obtained from diseased valves in situ, but tractionforces were not addressed in these studies.9

Furthermore, myxomatous MVIC appear to haveincreased expression of TGFb family memberreceptors, which, from the standpoint of our

Figure 5 Gene expression of ACTA2 (A), TGFb3 (B), and vimentin (C) under static control conditions (n ¼ 7), underbiaxial cyclic strain alone (n ¼ 5), with TGFb1 alone (n ¼ 8), and treated with both strain and TGFb1 (n ¼ 8). Data wasnormalized to GAPDH and expressed as fold difference from control. Bars denote statistically significant comparisons.

TGFb1 and cyclic strain in myofibroblastic differentiation 219

results, supports an increased traction forcegeneration leading to matrix compaction.2,34

Because pathogenic valve cell differentiation andmatrix remodeling in vivo is time dependent andheavily influenced by hemodynamic microenvi-ronment,35 we can’t yet establish the completepicture. Our results however suggest that changesin cell phenotype and matrix composition are notco-incident, but sequential and likely evolving ina complex feedback.

The results of our study indicate that 15% cyclicequibiaxial strain promotes a more fibroblasticphenotype in mildly diseased MVIC. The addition ofTGFb1 however caused significant increases inexpression of TGFb3 and aSMA mRNA in both static

Table 2 Fold gene expression changes formyofibroblast-like (ACTA2, TGFb3) and fibroblast-likephenotype (vimentin).

Gene Condition Fold changea

ACTA2 Cyclic strain 0.35 (0.07e1.64)ACTA2 TGFb1 1.12 (0.15e2.99)ACTA2 Cyclic strain

þ TGFb10.52 (0.24e2.00)

TGFb3 Cyclic strain 0.84 (0.31e2.29)TGFb3 TGFb1 1.80 (0.73e4.48)TGFb3 Cyclic strain

þ TGFb11.83 (1.18e2.85)

Vimentin Cyclic strain 0.53 (0.17e1.6)Vimentin TGFb1 2.47 (1.39e4.41)Vimentin Cyclic strain

þ TGFb10.93 (0.53e1.65)

a Fold change with upper and lower 95% confidenceintervals in parentheses.

and strained cultures. These findings suggest thatMVIC are responsive to both biomechanical andbiochemical stimuli. Under these experimentalconditions, strain and TGFb1 acted antagonisti-cally. The actual strain profiles in native mitralvalves may be more biaxial in nature, but heavilyheterogeneous.25 Early MVD is characterized byfocal lesions on the anterior leaflet in both dog andhuman, which suggests that local tissue strainpatterns may promote or protect valve cells fromdifferentiation. An analysis of mitral valve chordaeand anterior leaflets suggests that diseased tissueis biomechanically weaker,10 but direct correla-tions between local matrix composition and tissuemechanics are still lacking. These findings suggestthat resident cells may experience greater strainmagnitudes in diseased configurations. In aorticvalves, tissue strain magnitudes over 20% appear toinduce pathological cell differentiation and matrixremodeling, while 10e15% strains are protec-tive.36,37 Furthermore, Merryman et al17 has shownthat strain synergizes with TGFb1 stimulationresulting in additional TGFb stimulation and celldifferentiation. Uniaxial stretch was applied towhole leaflets in that study, which differs signifi-cantly from our studies.17 We employed equibiax-ial strain because it is a completely uniform strainfield, whereas uniaxial stretch imparts locallyvarying transverse strains. Our findings suggestthat 15% cyclic strain was protective against myo-fibroblastic differentiation of MVIC, but strainmagnitude and degree of anisotropy likely alsoaffect mitral valve homeostasis and pathogenesis,alone and in combination with biochemicalsignaling. Additional studies are therefore

220 A.S. Waxman et al.

required to dissect how each mode of stimulationinfluences MVIC phenotype and downstreammatrix remodeling.

In vivo studies have suggested that increasedcontractility may alter the progression of MMVD indogs with mild disease.37 Hypothetically, thealtered mechanical environment in this situationmay alter strain and therefore induce a moremyxomatous phenotype. However, cyclic strainimposed in our studies proved to be protective ofthe myxomatous phenotype. This suggests aninteresting possibility with regards to the severityof myxomatous disease seen in dogs of differentsizes and ages. Hypothetically, the variation in theclinical expression of the disease may be due notonly to genetic variation for the substrate of thedisease, but also the differences in strain on thevalve leaflets in different dogs may be a factor forconsideration in understanding the phenotypicvariation. In a sheep model of papillary muscletethering induced mitral valve stretch, tissueelongation and matrix remodeling with increasedexpression of aSMA was found primarily on thetensile loaded atrialis surface, which suggests thatmechanical forces independent of pharmacologicstimulation may also drive MV pathogenesis.35

It is important to note the limitations in ourstudy. First, we are studying a complex phenom-enon in a simplified model. Certainly, in vitromechanical studies cannot reproduce the multi-tude of factors present in vivo: however, our 3Dculture system enables decoupling and directtesting of the mechanical and biochemical alter-ations that are involved in the degenerativeprocess that cannot be studied in vivo. Further-more, while our analysis includes gene, protein,and functional readouts of the fibroblastic-myofibroblastic phenotype axis of MVIC, it is byno means an exhaustive list. Due to limitednumbers of cells isolated from diseased dogs andour desire to maintain low passage number, wewere unable to test the full comparison sets ofstrain versus TGFb1. Our experimental endpointsof 48 h for cell phenotype and seven days formatrix compaction are standard in the literature,but it is clear from Fig. 1 that cell inducedremodeling of collagen gels is time dependent.

Conclusions

These results support that TGFb1 induces myofi-broblastic differentiation (MVD phenotype) ofcanine MVIC in 3D culture, while 15% cyclic strainpromotes a more fibroblastic phenotype. 3Dculture of canine MVIC permits focused studies to

increase our knowledge on the mechanical andbiochemical changes that may be involved in thecellular phenotype of MVD.

Acknowledgments

We appreciate the statistical consultation and valu-able assistance of Dr. Mark Rishniw. We are verygrateful for the technical expertise and helpfuldiscussions with Dr. Eva Oxford. This study was sup-ported by a grant from the ACVIM subspecialty ofCardiology Resident Research Grant SupportaNational ScienceFoundationCAREERAward (CBETe0955172), the American Heart Association (ScientistDevelopment Grant #0830384N), the National Insti-tutes of Health (HL110328), and the Leducq Founda-tion (Project MITRAL).

Conflict of Interest

The authors have no conflict of interest.

References

1. Richards JM, Farrar EJ, Kornreich BG, Moı̈se NS, Butcher J.The mechanobiology of mitral valve function, degenera-tion, and repair. J Vet Cardiol 2012;14:47e58.

2. Rabkin E, Aikawa M, Stone JR, Fukumoto Y, Libby P,Schoen FJ. Activated interstitial myofibroblasts expresscatabolic enzymes and mediate matrix remodeling inmyxomatous heart valves. Circulation 2001;104:2525e2533.

3. Liu AC, Joag VR, Gotlieb AI. The emerging role of valveinterstitial cell phenotypes in regulating heart valvepathobiology. Am J Pathol 2007;171:1407e1418.

4. Hinton Jr RB, Lincoln J, Deutsch GH, Osinska H, Manning PB,Benson DW, Yutzey KE. Extracellular matrix remodeling andorganization in developing and diseased aortic valves. CircRes 2006;98:1431e1438.

5. Fox PR. Pathology of myxomatous mitral valve 1 disease inthe dog. J Vet Cardiol 2012;14:103e126.

6. Schneider PJ, Deck JD. Tissue and cell renewal in thenatural aortic valve of rats: an autoradiographic study.Cardiovasc Res 1981;15:181e189.

7. Han RIBA, Culshaw GJ, French AT, Else RW, Corcoran BM.Distribution of myofibroblasts, smooth muscle-like cells,macrophages, and mast cells, in mitral valve leaflets of dogswith myxomatous mitral valve disease. Am J Vet Res 2008;69:763e769.

8. Disatian S, Ehrhart 3rd EJ, Zimmerman S, Orton EC. Inter-stitial cells from dogs with naturally occurring myxomatousmitral valve disease undergo phenotype transformation. JHeart Valve Dis 2008;17:402e411.

9. Rabkin-Aikawa E, Farber M, Aikawa M, Schoen FJ, Source D.Dynamic and reversible changes of interstitial cell pheno-type during remodeling of cardiac valves. J Heart Valve Dis2004;13:841e847.

TGFb1 and cyclic strain in myofibroblastic differentiation 221

10. Barber JE, Kasper FK, Ratliff NB, Cosgrove DM, Griffin BP,Vesely I. Mechanical properties of myxomatous mitralvalves. J Thorac Cardiovasc Surg 2001;122:955e962.

11. Grashow JS, Sacks MS, Liao J, Yoganathan AP. Planar biaxialcreep and stress relaxation of the mitral valve anteriorleaflet. Ann Biomed Eng 2006;34:1509e1518.

12. Liao J, Yang L, Grashow J, Sacks MS. The relation betweencollagen fibril kinematics and mechanical properties in themitral valve anterior leaflet. J Biomech Eng 2007;129:78e87.

13. Oyama MA, Chittur SV. Genomic expression patterns ofmitral valve tissues from dogs with degenerative mitralvalve disease. Am J Vet Res 2006;67:1307e1318.

14. Zheng J, Chen Y, Pat B, Dell’Italia LA, Tillson M, Dillon AR,Powell PC, Shi K, Shah N, Denney T, Husain A, Dell’Italia LJ.Microarray identifies extensive downregulation of non-collagen extracellular matrix and profibrotic growth factorgenes in chronic isolated mitral regurgitation in the dog.Circulation 2009;119:2086e2095.

15. Disatian S, Orton EC. Autocrine serotonin and transforminggrowth factor beta 1 signaling mediates spontaneousmyxomatous mitral valve disease. J Heart Valve Dis 2009;18:44e51.

16. Jian B, Xu J, Connolly J, Savani RC, Narula N, Liang B,Levy RJ. Serotonin mechanisms in heart valve disease I:serotonin-induced up-regulation of transforming growthfactor-beta1 via G-protein signal transduction in aorticvalve interstitial cells. Am J Pathol 2002;161:2111e2121.

17. Merryman WD, Lukoff HD, Long RA, Engelmayr Jr GC,Hopkins RA, Sacks MS. Synergistic effects of cyclic tensionand transforming growth factor-beta1 on the aortic valvemyofibroblast. Cardiovasc Pathol 2007;16:268e276.

18. Butcher JT, Nerem RM. Valvular endothelial cells regulatethe phenotype of interstitial cells in co-culture: effects ofsteady shear stress. Tissue Eng 2006;12:905e915.

19. Ku CH, Johnson PH, Batten P, Sarathchandra P,Chambers RC, Taylor PM, Yacoub MH, Chester AH. Collagensynthesis by mesenchymal stem cells and aortic valveinterstitial cells in response to mechanical stretch. Car-diovasc Res 2006;71:548e556.

20. Gupta V, Tseng H, Lawrence BD, Grande-Allen KJ. Effect ofcyclic mechanical strain on glycosaminoglycan and proteo-glycan synthesis by heart valve cells. Acta Biomater 2009;5:531e540.

21. Balachandran K, Konduri S, Sucosky P, Jo H, Yoganathan AP.An ex vivo study of the biological properties of porcineaortic valves in response to circumferential cyclic stretch.Ann Biomed Eng 2006;34:1655e1665.

22. Lee AA, Delhaas T, McCulloch AD, Villarreal FJ. Differentialresponses of adult cardiac fibroblasts to in vitro biaxialstrain patterns. J Mol Cell Cardiol 1999;31:1833e1843.

23. Li Q, Muragaki Y, Hatamura I, Ueno H, Ooshima A. Stretch-induced collagen synthesis in cultured smooth muscle cellsfrom rabbit aortic media and a possible involvement ofangiotensin II and transforming growth factor-beta. J VascRes 1998;35:93e103.

24. Butcher JT, Nerem RM. Porcine aortic valve interstitial cellsin three-dimensional culture: comparison of phenotype

with aortic smooth muscle cells. J Heart Valve Dis 2004;13:478e485.

25. Sacks MS, Enomoto Y, Graybill JR, Merryman WD,Zeeshan A, Yoganathan AP, Levy RJ, Gorman RC,Gorman 3rd JH. In-vivo dynamic deformation of the mitralvalve anterior leaflet. Ann Thorac Surg 2006;82:1369e1377.

26. Gould RA, Chin K, Santisakultarm TP, Dropkin A,Richards JM, Schaffer CB, Butcher JT. Cyclic strainanisotropy regulates valvular interstitial cell phenotypeand tissue remodeling in three-dimensional culture. ActaBiomater in press.

27. Butcher JT, Barrett BC, Nerem RM. Equibiaxial strain stim-ulates fibroblastic phenotype shift in smooth muscle cells inan engineered tissue model of the aortic wall. Biomaterials2006;27:5252e5258.

28. Tranquillo RT, Durrani MA, Moon AG. Tissue engineeringscience: consequences of cell traction force. Cytotech-nology 1992;10:225e250.

29. Pomerance A, Whitney JC. Heart valve changes common toman and dog: a comparative study. Cardiovasc Res 1970;4:61e66.

30. Gupta V, Werdenberg JA, Blevins TL, Grande-Allen KJ.Synthesis of glycosaminoglycans in differently loadedregions of collagen gels seeded with valvular interstitialcells. Tissue Eng 2007;13:41e49.

31. El-Hamamsy I, Balachandran K, Yacoub MH, Stevens LM,Sarathchandra P, Taylor PM, Yoganathan AP, Chester AH.Endothelium-dependent regulation of the mechanicalproperties of aortic valve cusps. J Am Coll Cardiol 2009;53:1448e1455.

32. Grinnell F, Ho CH, Lin YC, Skuta G. Differences in theregulation of fibroblast contraction of floating versusstressed collagen matrices. J Biol Chem 1999;274:918e923.

33. Soini Y, Satta J, Maatta M, Autio-Harmainen H. Expression ofMMP2, MMP9, MT1-MMP, TIMP1, and TIMP2 mRNA in valvularlesions of the heart. J Pathol 2001;194:225e231.

34. Dal-Bianco JP, Aikawa E, Bischoff J, Guerrero JL,Handschumacher MD, Sullivan S, Johnson B, Titus JS,Iwamoto Y, Wylie-Sears J, Levine RA, Carpentier A. Activeadaptation of the tethered mitral valve: insights intoa compensatory mechanism for functional mitral regurgi-tation. Circulation 2009;120:334e342.

35. Smith KE,Metzler SA,Warnock JN. Cyclic strain inhibits acutepro-inflammatory gene expression in aortic valve interstitialcells. Biomech Model Mechanobiol 2010;9:117e125.

36. Balachandran K, Hussain S, Yap CH, Padala M, Chester AH,Yoganathan AP. Elevated cyclic stretch and serotonin resultin altered aortic valve remodeling via a mechanosensitive 5-HT(2A) receptor-dependent pathway. Cardiovasc Pathol2011 [Epub ahead of print].

37. Chetboul V, Lefebvre HP, Sampedrano CC, Gouni V,Saponaro V, Serres F, Concordet D, Nicolle AP,Pouchelon JL. Comparative adverse cardiac effects ofpimobendan and benazepril monotherapy in dogs with milddegenerative mitral valve disease: a prospective,controlled, blinded, and randomized study. J Vet InternMed 2007;21:742e753.

Available online at www.sciencedirect.com