Characterization of statically loaded tissue-engineeredmitral valve chordae tendineae

Yaling Shi, Ivan VeselyDepartment of Biomedical Engineering, ND20, Lerner Research Institute, The Cleveland Clinic Foundation, 9500Euclid Avenue, Cleveland, Ohio 44195

Received 5 December 2002; revised 17 September 2003; accepted 19 September 2003

Abstract: Chordae tendineae are essential to the properfunction of the mitral valve. Native chordae contain a densecollagenous core and an outer elastin sheath. We have beenusing the principle of directed collagen gel shrinkage tofabricate tissue-engineered mitral valve chordae. Becausethe microstructure of biologic tissues determines their me-chanical behavior, the morphology of collagen and elastin intissue-engineered chordae should mimic that of native chor-dae. The objective of this study, therefore, was to examinethe morphology of our tissue-engineered constructs in com-parison to native chordae. A collagen–cell suspension wascast into silicon rubber wells with microporous anchors atthe ends and cultured in an incubator. The anchors allowedshrinkage to occur only transverse to the long axis of thewells, thus creating highly aligned collagen fibril constructs.The collagen constructs were cultured for 8 weeks and char-acterized mechanically, histologically, and biochemically atdifferent culture time points. Histologic sections showedthat in all mature constructs collagen fibers were oriented

parallel to the long axis of the constructs. At the edge of thetissue collagen fibers were in general straight, whereas in themiddle of the tissue they were wavy. Transmission electronmicroscopy showed a progressive increase in the densityand longitudinal orientation of collagen fibrils with culturetime. Light and scanning electron microscopy showed thepresence of an elastin sheath around the collagen core. Im-munostaining demonstrated that smooth muscle cells differ-entiate during tissue development and TUNEL assayshowed that cells in the interior of the constructs undergoapoptosis. This study has demonstrated that collagen–cellconstructs, with material properties and microstructure sim-ilar to native mitral valve chordae, can be developed usingstatic culture. © 2004 Wiley Periodicals, Inc. J Biomed MaterRes 69A: 26–39, 2004

Key words: tissue-engineered collagen constructs; collagenfiber orientation; elastin sheath; smooth muscle cells

INTRODUCTION

The mitral valve consists of two leaflets, two papil-lary muscles, and multiple chordae tendineae thattransmit the forces generated by the closed valve tothe papillary muscles. To perform their intended func-tion, mitral valve chordae need to be elastic, strong,and fatigue resistant.1 The rupture of mitral valvechordae is reported to be the most common cause ofclinically significant mitral valve insufficiency.2–4 Ar-tificial chordae made from ePTFE (Teflon) have beenused to replace torn or dysfunctional chordae,5–8 butthey do not have appropriate mechanical propertiesand may also elicit a foreign-body reaction in therecipient.

In an effort to develop better materials for use in

chordal replacement, we turned to tissue-engineeringtechnologies. We used the principle of directed colla-gen gel shrinkage to fabricate mitral valve chordaewith good material properties.9 The principle of di-rected collagen gel shrinkage involves first mixingsolubilized collagen with the appropriate cells. Afterthe collagen–cell mixture is neutralized, soluble colla-gen reassembles into fibrils and a gel is created. Cellsbecome entrapped in the collagen gel and begin tointeract with the collagen fibrils. Tractional forces ap-plied to the collagen substrate during cell movementcontract and reorganize the collagen lattice. If the gelis mechanically constrained, collagen fibrils align inthe direction of constraint and highly aligned, com-pacted collagenous constructs can thus be fabri-cated.10–16

Collagen fibrils, the main constituent of mitral valvechordae, are oriented in the direction of maximumprincipal stress to serve the required mechanicalrole.17 Normal chordae tendineae also contain scat-tered elastin fibers within the inner core and two

Correspondence to: I. Vesely; e-mail: vesely@bme.ri.ccf.orgContract grant sponsor: United States Department of De-

fense; contract grant number: DAMD17-99-1-9475

© 2004 Wiley Periodicals, Inc.

distinct layers of elastic fibers and endothelium at theouter surface.18 Although some have reported thatremoval of the outer elastin sheath by enzymatic di-gestion does not significantly affect the mechanicalproperties of the chordae,19 Millington-Sanders et al.1

proposed that the elastic sheath would tend to restorethe collagen fibrils within the chordae back to theirwavy configuration upon relaxation of the appliedtension. This interaction between collagen and elastinhas been reported to exist in other valvular tissues.20

Because the microstructural arrangement of tissuecomponents determines the tissue’s mechanical be-havior, the morphology of the collagen and elastin inthe tissue-engineered chordae should mimic that ofnative chordae. The objective of this study, therefore,was to examine the morphology of tissue-engineeredconstructs and determine how closely the constructsmimic the native chordae.

MATERIALS AND METHODS

Collagen constructs were fabricated using the principle ofdirected collagen gel shrinkage, cultured up to 8 weeks, andcharacterized mechanically, histologically, and biochemi-cally at different culture time points.

Mold design

A rectangular chamber was constructed from silicone rub-ber in a 100-mm Petri dish. First, silicone rubber was pouredinto the Petri dish to a height of 5 mm. A block (36 � 18 �18 mm) was then placed in the dish and more silicone rubberwas poured into the dish to a further height of 6 mm andallowed to set overnight. The block was then removed,leaving a rectangular well. Strips of glass fiber (Millipore,Ireland) were wrapped around a stainless steel wire andplaced at the ends of the well to act as anchors for thecollagen gel. Prior to use, the silicone mold and the anchorswere steam sterilized at 121 psi for 15 min.

Primary cell culture

Neonatal rat aortic smooth muscle cells (SMCs) were iso-lated by the method outlined by Oakes et al.21 Briefly, seg-ments of aorta were incubated with 2 mL of type II collage-nase [2 mg/mL in Dulbecco’s modified eagle medium(DMEM)/nutrient mixture F12 (F12) (1:1) medium; Gibco,Auckland, NZ] for 10 min at 37°C to remove the endothe-lium. The explants were then washed with phosphate-buff-ered saline (PBS) several times, minced into small pieces,transferred onto a sterile Petri dish, and incubated in limitedvolumes of equal ratio of DMEM and F12, supplementedwith 10% fetal bovine serum (FBS; Life Technologies, GrandIsland, NY) at 37°C for 1 week to establish the primary

cultures. For subsequent passaging, the cells were detachedfrom the anchored dish by trypsinization with 1 cc of 0.05%fresh trypsin containing 0.2% ethylenediaminetetraaceticacid (EDTA) (Life Technologies, Grand Island, NY), sus-pended in the above medium, and centrifuged at 1500 rpm.The cell pellet obtained was resuspended, counted, andseeded in a 750mL Petri dish to amplify. Cells were stainedwith trypan blue, and a hemocytometer was used to deter-mine the cell density and viability. Culture medium waschanged twice a week.

Tissue culture

FBS and penicillin-streptomycin (pen-strep) were thawedand added to the medium (5 � DMEM/F12) to obtain asolution of 20% serum and 100 U/mL pen-strep. Sterileacid-soluble type I collagen (BD Biosciences, Bedford, MA,rat tail; 3.94 mg/mL, 0.02N acetic acid) was added to get aninitial concentration of 2.0 mg/mL, and the suspension wasbrought to physiological pH by the addition of 0.1N NaOH.The cells were detached from their culture dishes bytrypsinization, counted, centrifuged, and added to the col-lagen suspension to get a cell seeding density of 1.0 millioncells/mL. All mixing was done on ice. The collagen–cellsuspension was pipetted into the wells and incubated at37°C. Within several hours, a gel formed and attached to themicroporous holders at the ends of the wells. These holdersprevented longitudinal contraction and allowed shrinkageto occur only transverse to the long axis of the wells. Theconstructs were cultured for 2–8 weeks and culture mediumwas changed every 2 days.

Histology

When the cultures were terminated, constructs were cutfrom the anchors, fixed in 10% neutral buffered formalin(Sigma, St. Louis, MO) for 48 h, and histologically processed.Sections cut to 5 �m were mounted onto glass slides andstained with hematoxylin and eosin (H&E), Picrosirius red,Verhoeff’s, and Movat’s Pentachrome. Conventional andpolarized light microscopy of these sections was used toassess collagen, elastin, and cell morphology.

Fluorescent stain

Additional sections were cut as above, mounted on glassslides, and progressively rehydrated in xylene, 100% EtOHand 95% EtOH respectively, and then rinsed in tapwater.Slides were mounted in Vectashield with DAPI (VectorLabs, Burlingame, CA) and coverslipped. The sections wereilluminated with UV light, which reacts with DAPI to pro-duce blue light; the nuclei were then examined with a fluo-rescent microscope fitted with a 450-nm filter.

TISSUE-ENGINEERED MITRAL VALVE CHORDAE TENDINEAE 27

Mechanical testing

The mature constructs were cut from the anchors andmechanically tested using an Instron 8511 Plus series hy-draulic tensile testing load frame (Canton, MA) with a 5.0-lbload cell. The constructs were gripped with sandpaper-linedtissue clamps and mechanically tested in a bath of Hankssolution at 37°C. Specimens were first preconditioned at anextension rate of 4 mm/s for 10 cycles. The load/elongationcurves were usually visibly repeatable after three cycles.Failure tests were done as the last step and this curve wasused for analysis. The diameter of the constructs was mea-sured optically and the cross-section was assumed to becircular. Engineering stress was calculated by dividing theforce generated during extension by the initial cross-sec-tional area. Engineering strain, defined as extension dividedby the initial length, was used for estimating all materialparameters. Stiffness (modulus) was defined as the changein stress divided by the change in strain in the linear regionof the stress–strain curve. The extensibility measure wasobtained from the intercept of the stiffness slope.

Water content

To obtain their water content, the constructs wereweighed wet after gentle patting with bibulous paper (VWRScientific Products, West Chester, PA), lyophilized for 16 h,and then weighed dry. The water content was defined aswet weight minus dry weight divided by wet weight.

DNA assay

To measure DNA content, the tissues were first rehy-drated in 1 mL ammonium acetate (100 mM) and thendigested in proteinase K (Gibco, Galthersburg, MD) solution(10 mg/mL) in a 65°C water bath for 16 h. The completelysolubilized digests were boiled for 10 min to denature theproteinase K, the suspension was sonicated for 6 min on ice,and 200-�L aliquots of the digest were combined with 40 �Lof Hoechst dye no. 33258 (Polysciences Inc., Warrenton, PA),and evaluated in a fluorometer with excitation set to 356 nmand emission set to 458 nm (Hoefer Scientific InstrumentsDNA Fluorometer). The results were interpolated from astandard curve using calf thymus DNA. Because there areapproximately 6 pg of DNA per cell,22 the number of cellscan be estimated from DNA measurements.

Fastin Elastin assay

The Fastin Elastin assay is a commercially available kit formeasuring elastin (Biocolor Ltd., Ireland). Samples were putinto oxalic acid to extract elastin at 95°C for 60 min. All theoxalic acid extracts were pooled (supernatants collected) andconcentrated via centrifugation at 10,000 g for 10 min inmicrocentrifuge tubes fitted with membrane filters (MW

�15,000 cutoff). All �-elastin was retained because it has aMW of approximately 80,000. The oxalic acid-free proteinresidue above the filter was resuspended in distilled waterand the elastin was precipitated out by adding cold FastinPrecipitating Reagent. This purification step took place over-night in the refrigerator, after which the vial was centrifugedat 10,000–12,000 g for 10 min. All liquid was removed fromthe elastin pellet and the Fastin dye reagent was added. Thiswas followed by ammonium sulfate to make the elastin–dyecomplex precipitate out of solution. Samples were then cen-trifuged at 10,000 g for 10 min to isolate the elastin–dyecomplex and a Fastin Destain Reagent was added to bringthe dye into solution. One-hundred-microliter aliquots ofsamples and standards were put into a 96-well plate and readusing a microwell plate reader (SOFT MAX PRO) at 513 nm.

Scanning electron microscopy

Collagen constructs cultured for 8 weeks were mountedon holders, digested in 0.1N NaOH for 30 min at 75°C, andrinsed thoroughly with distilled water. The elastin remain-ing in the holders was frozen in 10% DMSO to �80°C for 1 hand lyophilized overnight. Dried samples were coated withgold and examined using scanning electron microscopy(SEM). Fastin elastin assay was also done on some samplesto verify that the structures being imaged contained onlypure elastin.

Transmission electron microscopy

Each sample was fixed in 1% glutaraldehyde and 4%formaldehyde, trimmed, and washed with cold sodium ca-codylate buffer (0.2M) three times for 5 min each. Sampleswere postfixed with 1% OsO4 for 60 min and stained with1% uranyl acetate for 60 min. Samples were dehydrated inethanol and propylene oxide/LX-112 (1:1) was added, setovernight, and changed to pure LX-112 medium. Sampleswere then embedded and observed via electron microscopy.

Immunostaining

Sections were stained for Von Willebrands factor, SM-�-actin, and prolyl-4-hydroxylase to determine the phenotypeof the cells in the constructs. Paraffin sections were rehy-drated and put in a microwave for 5 min in citrate buffer(pH � 6) for antigen retrieval (note: that this step wasomitted for SM-�-actin and prolyl-4-hydroxylase). Slideswere rinsed with 3% hydrogen peroxide for 5 min. Serumbuffer (1:10, 1000 �L PBS/100 �L goat serum) were appliedto the slides for 20 min to block any subsequent nonspecificbinding of goat-derived secondary antibody. Primary anti-bodies [SM-�-actin (1:200); prolyl-4-hydroxylase (1:500);Von Willebrands factor (1:50)] were added at an appropriatedilution and incubated for 60 min at 37°C. As a negativecontrol, goat serum/PBS was used to replace the primary

28 SHI AND VESELY

antibodies to check for nonspecific secondary antibody bind-ing. Secondary antibody (goat antimouse antibody, 1:200,2000 �L PBS/10 �L secondary antibody) were added andincubated for 30 min at 37°C. Sections were rinsed withavidin–biotin complex reagent for 30 min and then in DABsolution for 6–10 min at room temperature. Sections werethen washed and put in hematoxylin for 30 s, then rinsedwith tapwater for 5 min. Sections were then dehydrated andmounted with mounting media, covered with a coverslip,and viewed via a light microscope.

TUNEL assay

Cell apoptosis produces low-molecular-weight DNA frag-ments and single breaks (nicks) in high-molecular-weightDNA. These breaks in the DNA can be identified by labeling

the free 3’-OH termini with modified nucleotides in an en-zymatic reaction. In this commercially available kit (in situcell death detection kit, Roche Diagnostics GmbH, Ger-many), deoxynucleotidyl transferase (TdT) catalyzes the po-lymerization of fluorescein-labeled nucleotides to free 3’-OHDNA ends. Slides were dewaxed and the TUNEL reactionmixture added and incubated in humidity trays at 37°C for1 h, then mounted with Vectashield containing DAPI. Thefluorescein-labeled cells fluoresced green under fluorescencemicroscopy, while DAPI stained all the nuclei, producing ablue color. Green pixels that colocalized with the nucleitherefore denoted cells undergoing apoptosis.

RESULTS

Collagen gel contraction began in several hours andcontinued for up to 8 weeks of culture.9 At the end of

Figure 1. Images showing the process of fibroblasts-seeded collagen gel shrinkage over several days and weeks in culture.Passage 5 smooth muscle cells were mixed with serum and collagen solution to get a cell seeding density of 1M/mL and initialcollagen concentration of 2 mg/m: with 20% serum. Silicone rubber provides a surface to which the collagen/cell mixturecannot adhere. Holders at the ends entrap the collagen fibers and provide the uniaxial constraint that allows the constructsto shrink only transversely.

TISSUE-ENGINEERED MITRAL VALVE CHORDAE TENDINEAE 29

the culture period, the transparent gel became a dense,cylindrical construct (Fig. 1), with collagen fibersaligned along the long axis of the constructs. Theconstructs also had a continuous elastin sheath, a fea-ture typical of native chordae tendineae. To ourknowledge, the presence of an elastin sheath has notbeen reported in any other literature of collagen-basedconstructs. The mechanical properties of the con-structs demonstrated the typical nonlinear loadingcurve but were still at least an order of magnitudelower than those of native chordae tendineae.

Collagen core

Histology

Polarized light microscopy of picrosirius red-stained sections demonstrated that the organization ofcollagen fibers changed considerably during the cul-

ture period (Fig. 2). After 6 h or 2 days of culture, nocollagen fibrils were visible. At day 7, collagen fibrilsstill appeared randomly oriented. At day 15, however,most fibrils became aligned along the direction oftension but not densely compacted. At days 30 and 56,compacted collagen fibrils were fully oriented. At day56, they were more compacted than at day 30. Culturetime, therefore, is important for fibril orientation andcompaction.

We also observed that at the edge of the tissuecollagen fibers appeared straight, whereas in the mid-dle of the construct they exhibited a wavy pattern, acharacteristic of native chordae tendineae (Fig. 3). Thecrimp period in the middle was estimated to be 6.1 �1.82 �m.

Transmission electron microscopy

Transmission electron microscopy (TEM) images oflongitudinal sections revealed the characteristic

Figure 2. Picrosirius red-stained images showing the orientation and organization of collagen fibers in the constructs atdifferent culture times during the culture period. Note that collagen fiber organization changed greatly during tissuedevelopment.

30 SHI AND VESELY

banded appearance of collagen fibrils [Fig. 4(A, B)].There appeared to be progressive increase in fibrildensity and orientation with time, an observation con-sistent with the histological findings (Fig. 2). Images oflongitudinal sections taken from the central core of theconstructs suggested that constructs cultured for 8weeks [Fig. 4(B)] had more aligned and compactedfibrils than those cultured for 4 weeks [Fig. 5(A)].Images of the transverse sections demonstrated thatcollagen fibrils had roughly circular cross-sections asexpected, and constructs cultured for 8 weeks [Fig.4(D)] were denser than those cultured for 4 weeks[Fig. 4(C)]. This finding is consistent with the longitu-dinal sections. The fibrils in the older constructs werealso more circular and more evenly distributed, fea-tures indicative of greater longitudinal alignment.

Elastin sheath

Histology

Sections stained with Movat’s Pentachrome (Fig. 5)revealed the presence of elastin fibers within the col-lagen constructs after 4 weeks of culture. After 8weeks of culture, a continuous elastin sheath devel-oped around the dense collagen core. This is also acharacteristic of native mitral valve chordae. Becausethe constructs initially contained only cells and colla-

gen, the elastin sheath must have been synthesized bythe entrapped cells.

SEM

Fastin elastin assay showed that 90% of the residueremaining after both native chordae and the collagenconstructs were digested in 0.1N NaOH for 30 min at75°C was pure elastin. The structures observed underSEM are therefore continuous sheets of elastin. Nativechordae had a plate-like surface texture of the elastin[Fig. 6(A)], whereas our constructs were somewhatrougher and more irregular [Fig. 6(B)]. Higher-magni-fication images [Fig. 6(C)] showed the presence ofcrowded “pits,” roughly 10 �m in diameter, sugges-tive of imprints left by cells on the construct surface.At the ends of the constructs, within the glass fiberholders, elastin structure took the form of web-likesheets between adjacent glass fibers [Fig. 6(D)].

TEM

TEM images of longitudinal sections revealed thepresence of both microfibrils and patches of amor-phous elastin [Fig. 7(A)]. Images of longitudinal sec-tions confirmed that constructs cultured for 8 weekshad an elastin sheath around a collagen core [Fig.7(B)].

Figure 3. Picrosirius red-stained images of collagen fibers in a 4-week-old construct. Note that collagen fibers (A) at the edgewere straight whereas those (B) in the middle were more wavy. [Color figure can be viewed in the online issue, which isavailable at www.interscience.wiley.com.]

TISSUE-ENGINEERED MITRAL VALVE CHORDAE TENDINEAE 31

Figure 4. TEM images of longitudinal sections of the constructs cultured for (A) 4 weeks and (B) 8 weeks. Note that collagenfibrils in constructs cultured for 8 weeks were more aligned and compacted than those in the constructs cultured for 4 weeks,and also appeared somewhat longer. This was indicative of better alignment with the cutting plane. TEM images ofcross-sections showed that collagen fibers in constructs cultured for (D) 8 weeks were more dense, more circular, and alsomore evenly distributed than in those constructs cultured for (C) 4 weeks. This was consistent with the longitudinal sections.

32 SHI AND VESELY

Figure 5. Movat’s Pentachrome stain of elastin fibers scattered (A) within the interior of the tissue-engineered mitral valvechordae. These elastin fibers became evident only after 4 weeks of culture. Note the presence of a continuous elastin sheath(black) (B) at the edge of the construct. [Color figure can be viewed in the online issue, which is available at www.inter-science.wiley.com.]

Figure 6. SEM images of elastin isolated from (A) native chordae showed a plate-like surface texture, whereas elastin (B)from our constructs was somewhat rougher and more irregular. (C) Higher-magnification images showed the presence ofcrowded “pits,” roughly 10 �m in diameter, suggestive of imprints left by cells on the construct surface. (D) At the ends ofthe constructs, elastin showed web-like sheets between adjacent glass fibers.

TISSUE-ENGINEERED MITRAL VALVE CHORDAE TENDINEAE 33

Cells in the Constructs

Histology

Smooth muscle cells in the interior of the constructshad an elongated, bipolar shape, typically observed intissues, rather than the single broad lamellae seen onculture dishes. In constructs cultured for 4 weeks, cellsappeared to follow the contours of the collagenfibrils—at the edge of the tissue they were elongatedand straight, whereas in the middle of the tissue theywere much more wavy (Fig. 8).

Fluorescent staining

Fluorescent images showed large amounts ofsmooth muscle cells around the surface of the con-structs (Fig. 9), even though the cells were originallydispersed evenly through the gel. The number of cellsseen on the surface of the collagen constructs in-creased progressively with culture time, eventuallyforming a continuous sheet, and multiple layers. Theupper cell layers frequently detached during han-dling, indicating weak cell interaction. Cells accumu-lated at the surface of constructs and appeared tochange their morphology over time. At day 14, cellswere all bipolar, just like the cells in the interior. Atday 21, however, cells at the surface became rounderwhereas those in the middle remained elongated. Thissuggested that cells on the surface might have a phe-notype different from those in the interior.

Figure 7. (A) TEM images of longitudinal sections revealedthe presence of both patches of (*) amorphous elastin and(**) microfibrils. (B) Low-magnification TEM images alsodemonstrated that constructs cultured for 8 weeks had anelastin sheath around the collagen core.

Figure 8. Movat’s Pentachrome stain of cells in tissue-engineered mitral valve chordae. Note that cells appeared to followthe contours of the collagen fibrils. (A) At the edge of the tissue they were elongated and straight, whereas (B) in the middleof the tissue they were more wavy. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

34 SHI AND VESELY

Immunostaining

Cells stained negative for factor VIII, indicating thatthere were no endothelial cells present (Fig. 10). Im-munostaining of collagen constructs for SM-�-actinand prolyl-4-hydroxylase showed that most of thecells in the middle of the constructs were of the con-tractile phenotype whereas most of the cells at thesurface were of the synthetic phenotype.

DNA assay

During the first 3 weeks, DNA content increasedalmost linearly, reach a steady phase between 3 and 6weeks of culture, and then decreased almost linearlyafterward (Fig. 11). This suggests that cells prolifer-ated initially and then began to die off.

TUNEL assay

Cells in the constructs underwent apoptosis duringtissue development (Fig. 12). The percentage of apo-ptotic cells started at 20%, then decreased with time,but progressively increased again after 4 weeks ofculture (Fig. 13).

Biophysical properties

The water content of the SMC-seeded constructswas similar to that of native chordae.9 During mechan-ical testing, all constructs showed pronounced hyster-esis and progressive elongation during precondition-ing and the nonlinear stress–strain curves typical ofcollagenous tissue (Fig. 14). The constructs had anextensibility of 9.25% � 2.64% (strain to lock-up,mean � SD), a modulus of 5.33 � 1.37 MPa (measuredin the posttransition region), and a failure strength of1.13 � 0.45 MPa. While the extensibility of our con-structs is comparable to that of native chordae (15.4 �2.2% strain for human basal chordae; 9.2 � 1.1% strainfor human marginal chordae), the stiffness and failurestrength of our constructs still need to be improved byat least an order of magnitude.

DISCUSSION

The structural matrix of mitral valve chordae mustsatisfy demanding mechanical requirements to with-stand the large repetitive tensile forces encounteredwithin the left ventricle. It has been suggested that thewavy arrangement of collagen fibers surrounded by

Figure 9. DAPI staining of nuclei in the constructs at different culture times. The number of cells seen on the surface of thecollagen constructs increased progressively with culture time, eventually forming a continuous sheet and multiple layers.Note that the cells that accumulated at the surface of constructs appeared to change their morphology from elongated torounded.

TISSUE-ENGINEERED MITRAL VALVE CHORDAE TENDINEAE 35

an elastic sheath is a configuration that is well adaptedto the cyclic stress to which the chordae are continu-ously subjected.1 When a force is applied, the initialresponse of the chordae is a straightening of the col-lagen crimp and rapid stiffening. Any further strainoccurs through a stretching of the collagen fibril net-

work. Once the tension is released, the collagen fibrilsreturn to their wavy configuration. It is presumed that,as in other tissues, elastin somehow returns the colla-gen fibers back to their resting configuration.20

Collagen is the primary load-bearing component ofchordae and, accordingly, is arranged along the linesof principal stress. Why chordae have a greater crimpin the center than at the edges is at present unknown.Coincidentally, our constructs also had a greater wav-iness in the center than at the edges. We can onlyspeculate that the reason for the different collagencrimp across the thickness of our constructs resultedfrom a variation of tensile force during culture. Be-cause the constructs are narrower in the center than atthe ends, the distance between anchors along the edgeof the constructs is greater than that along the centralfibers. It is therefore conceivable that the outer fibersexperience greater strains than those in the core andwould thus be less crimped (Fig. 15).

It is well known that SMCs display remarkable plas-ticity in terms of differentiation, proliferation, andmotility. Vascular smooth muscle cells have beenshown to progress from a proliferative phase to a

Figure 10. Immunostaining for cell phenotype (original uncropped magnification, 20�). (A, D, and G) Positive controls forfactor VIII, SM-�-actin, and prolyl-4-hydroxylase, respectively. The control tissues were (A) vasa-vasora in artery, (D)arterioles in skin, and (G) epidermis. Cells stained negative for factor VIII, indicating that there were no endothelial cellspresent in the constructs, either (B) on the surface or (C) in the interior. Immunostaining for SM-�-actin showed that only afew cells on the surface were (E)of the contractile phenotype whereas most of the cells in the middle of the constructs were(F) of the contractile phenotype. Immunostaining for prolyl-4-hydroxylase showed that most of the (H) cells on the surfacewere of the synthetic phenotype whereas few of the (I) cells in the interior were of the synthetic phenotype. [Color figure canbe viewed in the online issue, which is available at www.interscience.wiley.com.]

Figure 11. During the first 3 weeks, DNA content in-creased almost linearly, reached a steady state between 3and 6 weeks of culture, and then decreased almost linearlyafterward.

36 SHI AND VESELY

synthetic to a quiescent contractile phase.23, 24 Thisphenotypic change is reversible, and such “dediffer-entiation” has been shown to precede the onset ofproliferation. The change in phenotype is influencedby the presence of extracellular matrix proteins andthe mechanical environment.25 Confluent cultures ofSMCs have been reported to have higher levels ofmuscle-specific proteins than subconfluent culturesand therefore may be representative of the contractilephenotype.26 A 3D matrix promotes the differentiated(contractile) SMC phenotype.25

In our system, cells in the contractile phenotypeobtained from confluent cultures were resuspendedand seeded into collagen gels. Because the connectionsbetween cells and ECM were totally destroyed bydigestion, the cells may have changed their phenotypeto the synthetic state upon being seeded in the con-structs. They were thus primed to produce, as well asrearrange, the extracellular matrix that surrounds

them. The process of suspending the cells prior toseeding may have also induced the high level of initialapoptosis. As the tissues matured, the cells likely con-verted to the contractile phenotype with a decrease inproliferative and synthetic activity. Indeed, our immu-nostaining demonstrated that most cells in the middleof the 8-week-old constructs were in the contractilephenotype. Static tension in the constructs progres-sively increased with compaction and collagen fiberorientation, likely stimulating SMC’s phenotype con-version. This finding is consistent with previous re-ports of a 3D matrix promoting the contractile pheno-type.25

Our results also showed that cells progressivelyaccumulated on the surface of the constructs and grad-ually formed multiple cell layers. The weak interactionbetween the cell layers suggested that cells in the

Figure 12. TUNEL assay of the collagen constructs showing apoptotic cells staining green. [Color figure can be viewed inthe online issue, which is available at www.interscience.wiley.com.]

Figure 13. The percentage of apoptotic cells started at 20%,then decreased with time, but progressively increased againafter 4 weeks of culture. Error bars represent the standarddeviation of 20 images analyzed at each time point.

Figure 14. Failure test, after 10 cycles of preconditioning, ofthe SMC-seeded (1 million cells/mL) constructs with aninitial collagen concentration of 2 mg/mL, cultured for 56days. Note the nonlinear stress–strain curve and the meth-ods used to extract material parameters.

TISSUE-ENGINEERED MITRAL VALVE CHORDAE TENDINEAE 37

upper layer of the constructs were in much less ten-sion than in the interior. Cells that migrated to thesurface therefore proliferated and remained in thesynthetic phenotype. Indeed, our immunostainingshowed that most of the cells on the surface of theconstruct were of the synthetic phenotype. Elastin istherefore being synthesized by these cells and an elas-tin sheath eventually develops. Cells at the edge of theconstruct can absorb nutrients more readily and pro-liferate more quickly—perhaps the development ofthe outer elastin sheath is driven by multiple factors.

The similarity between the artificial chordae andnative chordae may result from the fact that they bothdevelop in the presence of tensile loads. In artificialchordae, however, the tension is static, whereas nor-mal mitral valve chordae develop under dynamicloads. Future work will therefore examine the role ofdynamic loading in a bioreactor that can simulate theloading conditions that exist in vivo.

The authors thank the U.S. Department of Defense for thefinancial support of this work under Grant DAMD17-99-1-9475. The authors are grateful to K. Jane Grande-Allen andEd Barber of their laboratory for assistance in harvestingcells, histology assays, and materials testing.

References

1. Millington-Sanders C, Meir A, Lawrence L, Stolinski C. Struc-ture of chordae tendineae in the left ventricle of the humanheart. J Anat 1998;192:573–581.

2. Akins CW, Hilgenberg AD, Buckley MJ, Vlahakes GJ, Torchi-ana DF, Daggett WM, et al. Mitral valve reconstruction versusreplacement for degenerative or ischemic mitral regurgitation.Ann Thoracic Surg 1994;58:668–675; discussion 675–676.

3. Hanson TP, Edwards BS, Edwards JE. Pathology of surgicallyexcised mitral valves. One hundred consecutive cases. ArchPathol Lab Med 1985;109:823–828.

4. Yacoub M, Halim M, Radley-Smith R, McKay R, Nijveld A,Towers M. Surgical treatment of mitral regurgitation caused byfloppy valves: repair versus replacement. Circulation 1981;64(2, Pt 2):II210–II216.

5. Cochran RP, Kunzelman KS. Comparison of viscoelastic prop-erties of suture versus porcine mitral valve chordae tendineae.J Card Surg 1991;6:508–513.

6. David TE. Replacement of chordae tendineae with expandedpolytetrafluoroethylene sutures. J Card Surg 1989;4:286–290.

7. David TE, Bos J, Rakowski H. Mitral valve repair by replace-ment of chordae tendineae with polytetrafluoroethylene su-tures. J Thorac Cardiovasc Surg 1991;101:495–501.

8. Zussa C, Polesel E, Rocco F, Galloni M, Frater RW, Valfre C.Surgical technique for artificial mitral chordae implantation.J Card Surg 1991;6:432–438.

9. Shi Y, Vesely I. Fabrication of tissue engineered mitral valvechordae using directed collagen gel shrinkage. Tissue Eng2003;9(6):1233–1242.

10. Bell E, Ivarsson B, Merrill C. Production of a tissue-like struc-ture by contraction of collagen lattices by human fibroblasts ofdifferent proliferative potential in vitro. Proc Natl Acad SciUSA 1979;76:1274–1278.

11. Grinnell F. Fibroblasts, myofibroblasts, and wound contrac-tion. J Cell Biol 1994;124:401–404.

12. Guidry C, Grinnell F. Studies on the mechanism of hydratedcollagen gel reorganization by human skin fibroblasts. J CellSci 1985;79:67–81.

13. Harris AK, Stopak D, Wild P. Fibroblast traction as a mecha-nism for collagen morphogenesis. Nature 1981;290(5803):249–251.

14. Stopak D, Harris AK. Connective tissue morphogenesis byfibroblast traction. I. Tissue culture observations. Dev Biol1982;90:383–398.

15. Tranquillo RT, Durrani MA, Moon AG. Tissue engineeringscience: consequences of cell traction force. Cytotechnology1992;10:225–250.

16. Barocas VH, Tranquillo RT. An anisotropic biphasic theory oftissue-equivalent mechanics: the interplay among cell traction,

Figure 15. Schematic diagram of the likely stress distribution model within the constructs. Because the distance betweenanchors along the edge of the constructs is greater than that along the central fibers, there is an uneven strain or stressdistribution: low in the center and increasing toward the edges. This may be the reason for the lower collagen fiber crimp atthe edges.

38 SHI AND VESELY

fibrillar network deformation, fibril alignment, and cell contactguidance. J Biomech Eng 1997;119:137–145.

17. Hiltner A, Cassidy JJ, Baer E. Mechanical properties of biolog-ical polymers. Annu Rev Mater Sci 1985;15:455–482.

18. Fenoglio JJ Jr, Tuan Duc P, Wit AL, Bassett AL, Wagner BM.Canine mitral complex. Ultrastructure and electromechanicalproperties. Circ Res 1972;31:417–430.

19. Lim KO, Boughner DR. Morphology and relationship to exten-sibility curves of human mitral valve chordae tendineae. CircRes 1976;39:580–585.

20. Vesely I. The role of elastin in aortic valve mechanics. J Bio-mech 1998;31:115–123.

21. Oakes BW, Batty AC, Handley CJ, Sandberg LB. The synthesis ofelastin, collagen, and glycosaminoglycans by high density pri-mary cultures of neonatal rat aortic smooth muscle. An ultrastruc-tural and biochemical study. Eur J Cell Biol 1982;27:34–46.

22. Labarca C, Paigen K. A simple, rapid, and sensitive DNA assayprocedure. Anal Biochem 1980;102:344–352.

23. Thyberg J, Blomgren K. Phenotype modulation in primarycultures of rat aortic smooth muscle cells. Effects of drugs thatinterfere with the functions of the vacuolar system and thecytoskeleton. Virchows Arch B Cell Pathol Incl Mol Pathol1990;59:1–10.

24. Thyberg J, Hedin U, Sjolund M, Palmberg L, Bottger BA.Regulation of differentiated properties and proliferation ofarterial smooth muscle cells. Arteriosclerosis 1990;10:966 –990.

25. Li X, Tsai P, Wieder ED, Kribben A, Van Putten V, Schrier RW,et al. Vascular smooth muscle cells grown on Matrigel. Amodel of the contractile phenotype with decreased activationof mitogen-activated protein kinase. J Biol Chem 1994;269:19653–19658.

26. Shirinsky VP, Birukov KG, Koteliansky VE, Glukhova MA,Spanidis E, Rogers JD, et al. Density-related expression ofcaldesmon and vinculin in cultured rabbit aortic smooth mus-cle cells. Exp Cell Res 1991;194:186–189.

TISSUE-ENGINEERED MITRAL VALVE CHORDAE TENDINEAE 39