Biomaterials 24 (2003) 2523–2532

A novel bioreactor for the dynamic flexural stimulationof tissue engineered heart valve biomaterials

George C. Engelmayr Jr.a, Daniel K. Hildebranda, Fraser W.H. Sutherlandb,John E. Mayer Jr.b, Michael S. Sacksa,*

aEngineered Tissue Mechanics Laboratory (ETML), McGowan Institute for Regenerative Medicine, Department of Bioengineering,

University of Pittsburgh, 100 Technology Drive, Room 250, Pittsburgh, PA 15219, USAbDepartment of Cardiovascular Surgery, Children’s Hospital Boston, Harvard Medical School, Boston, MA, USA

Received 24 May 2002; accepted 20 January 2003

Abstract

Dynamic flexure is a major mode of deformation in the native heart valve cusp, and may effect the mechanical and biological

development of tissue engineered heart valves (TEHV). To explore this hypothesis, a novel bioreactor was developed to study the

effect of dynamic flexural stimulation on TEHV biomaterials. It was implemented in a study to compare the effect of uni-directional

cyclic flexure on the effective stiffness of two candidate TEHV scaffolds: a non-woven mesh of polyglycolic acid (PGA) fibers, and a

non-woven mesh of PGA and poly l-lactic acid (PLLA) fibers, both coated with poly 4-hydroxybutyrate (P4HB). The bioreactor

has the capacity to dynamically flex 12 rectangular samples (25� 7.5� 2mm) under sterile conditions in a cell culture incubator.Sterility was maintained in the bioreactor for at least 5 weeks of incubation. Flexure tests to measure the effective stiffness in the

‘‘with-flexure’’ (WF) and opposing ‘‘against-flexure’’ (AF) directions indicated that dynamically flexed PGA/PLLA/P4HB scaffolds

were approximately 72% (3 weeks) and 76% (5 weeks) less stiff than static controls (po0:01), and that they developed directionalanisotropy by 3 weeks of incubation (stiffer AF, po0:01). In contrast, both dynamically flexed and static PGA/P4HB scaffoldsexhibited a trend of decreased stiffness with incubation, with no development of directional anisotropy. Dynamically flexed PGA/

P4HB scaffolds were significantly less stiff than static controls at 3 weeks (po0:05). Scanning electron microscopy revealed signs ofheterogeneous P4HB coating and fiber disruption, suggesting possible explanations for the observed mechanical properties. These

results indicate that dynamic flexure can produce quantitative and qualitative changes in the mechanical properties of TEHV

scaffolds, and suggest that these differences need to be accounted for when comparing the effects of mechanical stimulation on the

development of cell-seeded TEHV constructs.

r 2003 Elsevier Science Ltd. All rights reserved.

Keywords: Tissue engineered heart valve; Dynamic mechanical stimulation; Flexure; Bioreactor

1. Introduction

Bioreactors have been developed for the dynamicmechanical stimulation of tissue engineered cardiovas-cular constructs, including vascular grafts [1,2], myo-cardial patches [3], and heart valves [4–6]. These devicestypically rely on pulsatile flow to generate a complexbiomechanical environment resembling in vivo condi-tions, and have been demonstrated to promote both thedevelopment of mechanical strength [1,2,7], and themodulation of cellular function [8] within these tissue

engineered constructs. While these devices have shownpromise in the development of functional tissue replace-ments [1,7], they present several drawbacks whenapplied to the study of fundamental biomechanicalphenomena, including: small sample capacity, anatomi-cal sample geometry, and coupled mechanical stimuli.Few bioreactors have been designed for the explicitpurpose of studying the individual effect of specificmodes of mechanical stimulation on engineered cardi-ovascular tissues, such that candidate scaffolds andconstructs can be systematically evaluated [9–12]. Suchdevices should be designed to provide a user-defined,simple mode of mechanical stimulation, offer a sufficientsample capacity for statistically significant comparisonsat multiple time points, and accommodate a simple

*Corresponding author. Tel.: +1-412-235-5146; fax: +1-412-235-

5160.

E-mail address: msacks@pitt.edu (M.S. Sacks).

0142-9612/03/$ - see front matter r 2003 Elsevier Science Ltd. All rights reserved.

doi:10.1016/S0142-9612(03)00051-6

sample geometry amenable to subsequent mechanicaltesting. As is the case for most tissue engineeringbioreactors, these devices should allow for the cultureof engineered tissue samples under sterile, physiological(e.g., 37�C and 5% CO2) conditions for the duration ofthe experiment, and should be relatively easily to cleanand maintain.In our laboratory we are interested in studying the

effect of dynamic mechanical stimulation on the devel-opment of engineered heart valve tissues. Our workincludes both the pragmatic evaluation of candidatescaffold materials, as well as studies to investigate thefundamental mechanisms by which mechanical forcesstimulate tissue development. Moreover, our long-termgoal is that a rational basis for tissue engineering designmay eventually be derived from such studies, allowingtissue structure and function to be predicted from cell,scaffold, media, and environmental conditions.In vivo, heart valves are subjected to a unique

combination of mechanical stimuli, including flexure,shear stress, and tension [13]. Whereas the effects ofshear stress and tension on the development ofengineered tissues have been explored to an extent[9,12], the independent effect of dynamic flexuralstimulation on TE cardiovascular constructs has notbeen studied to date. Because flexure represents a majormode of deformation in heart valve leaflets [14], wesought to develop a device to study the independenteffect of flexure on tissue engineered heart valve (TEHV)scaffolds and constructs.In the current study, our goals were three-fold: (1) to

develop a bioreactor to provide cyclic flexural stimula-tion, (2) to demonstrate the operation of the bioreactorand sterility maintenance, and (3) to evaluate the effectsof uni-directional cyclic flexure on the effective stiffnessof bioresorbable polymeric scaffolds which have beenused extensively in the tissue engineering of heart valves.For the last goal, we used a non-woven polyglycolic acid(PGA) fiber mesh coated with poly 4-hydroxybutyrate(P4HB), and a non-woven, 50:50 blend, mesh of PGAand poly l-lactic acid (PLLA) fibers coated with P4HB.

2. Materials and methods

2.1. Bioreactor design

The bioreactor was designed using Solidworks 3DCAD software (Solidworks Corp., Concord, MA), andthe parts were fabricated by Apollo Precision, Inc.(Plymouth, MN). The structural elements of the devicewere machined from polysulfone, chosen for its excellentthermal and chemical stability, and abrasion-resistantacrylic, which provides good optical transparency.Neoprene gaskets were fitted between all critical jointsin order to reduce the possibility of microbial contam-

ination, and the bioreactor was assembled using 18–8stainless-steel screws (McMaster-Carr Supply Co.,Cleveland, OH).The bioreactor consists of two identical chambers

(127mm� 101.6mm), each containing 6 culturewells (25.4mm diameter, 16mm deep). Situated withineach well are four stainless-steel ‘‘stationary posts’’arranged orthogonally around a central channel in thefloor of the culture well (1.9mm deep) (Fig. 1). Thedevice can thus accommodate a total of 12 rectan-gular samples (maximum dimensions approximately25mm� 7.5mm� 2mm), with each sample beingpositioned between the four stationary posts, orthogo-nal to the central channel.

C

DB

A

Fig. 1. Schematic diagram of a single chamber of the bioreactor. The

bioreactor consists of two parallel chambers coupled via a cross-arm to

a centrally positioned linear actuator (Fig. 2). Each chamber contains

six culture wells (C), within which four ‘‘stationary posts’’ are

positioned orthogonally around a central channel. The linear

displacement of the actuator (not shown) is translated into uni-

directional or bi-directional three-point flexure via the cross-arm (not

shown) and the arm (A). The arm connects to the cross-arm at (A),

bifurcates before penetrating the wall of the chamber, and terminates

in six fingers (D). Two ‘‘flexure pins’’ can be inserted through each

finger, protruding into each culture well channel such as to bracket the

center of a rectangular sample. The interior of the chamber (B) is

protected from contamination by a lid (not shown).

G.C. Engelmayr Jr. et al. / Biomaterials 24 (2003) 2523–25322524

Dynamic flexural stimulation can be applied to each ofthe samples in the form of cyclic three-point flexure byan environmentally sealed linear actuator (UltraMotion,Mattitick, NY). The piston of the actuator is rigidlycoupled to a cross-arm in the form of a T-junction.Conversely, the cross-arm is rigidly coupled to the arm ofeach chamber (Fig. 1). Each arm bifurcates and extendsinto a chamber (two penetrations per chamber). Botharms terminate in six fingers through which ‘‘flexurepins’’ can be inserted to bracket the rectangular samplesin the middle. Therefore, each sample can be subjected touni-directional or bi-directional three-point flexure.Frequency, amplitude, acceleration, and decelerationprofiles can be developed using Windows-based SiProgrammer software (Applied Motion Products, Wat-sonville, CA). The structural elements of the device canbe cold gas sterilized by ethylene oxide, and the entiredevice was designed to be operated inside of a standardhumidified incubator (Fig. 2).

2.2. Culture medium

The culture medium used in these experiments wasDulbecco’s Modified Eagle’s Medium with 4.5 g/lglucose and l-glutamine supplemented with 10% fetalbovine serum (BioWhittaker, Walkersville, MD). Anti-

biotics were excluded in order to assess the sterility ofthe bioreactor.

2.3. Scaffolds

Scaffold materials consisted of a non-woven mesh ofPGA fibers (Albany International Research, Mansfield,MA) dip-coated with P4HB (Tepha, Inc., Cambridge,MA), and a non-woven 50:50 blend mesh of PGA andPLLA fibers (Albany International Research, Mans-field, MA) dip-coated with P4HB. The PGA and PGA/PLLA scaffolds had an approximate fiber diameter of0.012–0.015mm and density of 69mg/ml. Rectangularscaffold samples were cut to size (approximately25� 7.5� 2mm) and dipped briefly into a solution ofP4HB in tetrahydrofuran (1%wt/vol), resulting in aP4HB coating following solvent evaporation. As de-scribed previously [7], P4HB is a bioresorbable thermo-plastic that allows for scaffolds to be molded into anyshape. Scaffolds were cold gas sterilized with ethyleneoxide prior to use.

2.4. Flexure experiments

A total of 60 scaffold samples were prepared forevaluation in this study (30�PGA/P4HB, 30�PGA/

Fig. 2. Bioreactor operating inside a standard cell culture incubator at 37�C and 5% CO2. The bioreactor consists of two parallel chambers (Fig. 1)

secured to a base plate and coupled via a cross-arm to a centrally positioned linear actuator.

G.C. Engelmayr Jr. et al. / Biomaterials 24 (2003) 2523–2532 2525

PLLA/P4HB). Two separate runs of the bioreactor wererequired to test all of the samples. In the first run, PGA/P4HB scaffold samples were loaded into the culturewells of the bioreactor under aseptic conditions(n ¼ 12). Control scaffold samples were loaded intoidentical culture wells maintained under static condi-tions inside a large Petri dish (n ¼ 12). In addition,samples of the virgin scaffold material (time=0) wereretained for flexure testing and scanning electronmicroscopy (SEM) (n ¼ 6).Approximately 6ml of culture medium was added to

each of the bioreactor and static control culture wells.The linear actuator was programmed to flex the scaffoldsamples at a rate of 1 cycle per second (1Hz) in order toemulate the cardiac cycle. The center of each scaffoldsample was displaced 6.35mm in one direction, corre-sponding to a flexure angle of approximately 62�. Thedirection of flexure is referred to henceforth as ‘‘with-flexure’’ (WF), and the opposing direction is referred toas ‘‘against-flexure’’ (AF). After loading the bioreactor,the entire device was placed inside of a humidifiedincubator (HERAcell, Kendro Laboratory Products,Newtown, CT) operating at 37�C and 5% CO2. Theculture medium was not changed during the course ofthe run.After 1 week (n ¼ 6) and 3 weeks (n ¼ 6) of dynamic

flexure the scaffold samples were removed from thebioreactor under aseptic conditions to perform flexuretesting and SEM. Corresponding control samples wereremoved from the static culture wells at each time point.In the second bioreactor run, PGA/PLLA/P4HB scaf-fold samples were loaded and incubated in the samemanner as the PGA/P4HB scaffolds. After 3 weeks(n ¼ 6) and 5 weeks (n ¼ 6) of dynamic flexure thePGA/PLLA/P4HB scaffold samples were removed fromthe bioreactor, as well as the corresponding staticcontrol samples.

2.5. Sterility evaluation

The macroscopic appearance of the culturemedium was checked regularly during the course ofthe bioreactor runs for signs of contamination(i.e., cloudiness). At each time point evaluated in thestudy, upon removing the scaffold sample fromthe bioreactor or static control culture well, the culturemedium was aseptically transferred from the culturewell into a sterile T-25 flask (Becton DickinsonLabware, Franklin Lakes, NJ). The culture mediumwas then inspected for signs of contamination usingan inverted microscope (Micromaster Inverted Micro-scope, Fisher Scientific, Pittsburgh, PA). Theculture medium was retained in the incubator at 37�Cand 5% CO2 for 1 week and re-checked for signs ofcontamination.

2.6. Effective stiffness testing

The device and general testing methods used by ourlab to measure the effective stiffness of tissue sampleshave been described previously [15]. In brief, rectangularscaffold samples were measured, marked with 5–10small black dots (B0.3mm) along one long edge using apermanent marker, and then subjected to three-pointflexure by a calibrated flexure bar of known stiffness.The displacement of the scaffold sample was recordedon an S-VHS recorder using a high resolution CCDcamera (Fig. 3).In contrast to biological tissue, non-woven fiber mesh

scaffolds are relatively compressible. In the currentstudy an imaging method was used in favor of calipersto measure the thickness of the samples. A digital imageof each sample was captured from the S-VHS video andcalibrated using SigmaScan Pro (SPSS, Inc., Chicago,IL). The thickness of the sample was measured bytracing a line across the thickness of the sample at fiveuniformly spaced locations along its length (Fig. 4). Themean value of the line length at the five locations wasused as the thickness in the data analysis calculations to

Fig. 3. Effective stiffness testing video images. Recording starts with

the sample in the horizontal, un-deformed reference state (A). During

data analysis, the positions of the flexure bar (I), reference rod (II), and

sample markers (III) are recorded throughout the entire breadth of the

three-point flexure deformation (B).

G.C. Engelmayr Jr. et al. / Biomaterials 24 (2003) 2523–25322526

obtain the effective stiffness of the scaffold samplematerial.The data analysis procedure used here extends on our

previously described methods, allowing for large defor-mations, and thus warrants a detailed description. Theentire data analysis process can be decomposed intothree individual steps: (1) calibration, (2) tracking, and(3) calculation. In the calibration step, the number ofpixels per millimeter is determined for the video data,such that the data can be coupled to the physicaldimensions of the sample. An image is captured fromthe first few seconds of the video data, during whichtime a calibration grid is positioned within the viewablearea. This calibration image is analyzed using SigmaS-can Pro (SPSS, Inc.), and the resolution (pixels/mm) iscalculated and recorded. In addition, the coordinates ofthe two posts (Fig. 3) are determined. The resolutionand post coordinates are used in the subsequent dataanalysis steps.In the second step, the coordinates of the sample

markers (small black dots marked along the long edge ofthe sample), the flexure bar marker, and the referencerod marker are tracked using a custom C program. Thevideo tape is cued to the beginning of the flexure test, theinitial location of the sample, flexure bar, and referencerod markers are defined by the user, and then the videoand tracking procedure are started simultaneously, suchthat the marker coordinates are tracked and recordedthroughout the entire range of one loading cycle. Theprogram computes the applied load by multiplyingthe displacement of the flexure bar marker from thereference rod marker by a force constant, a numberdetermined beforehand by calibrating the flexure bar asdescribed previously [15]. The time, load, displacement,and marker coordinate data are all stored in a file that isused during the calculation step of the analysis.

In the third and final step of data analysis, the datafile generated in the tracking step is processed using acustom Matlab program. Based on the resolution, postcoordinates, load, displacement, and marker coordi-nates, the program calculates the applied moment, M ;and the resulting change in curvature, Dk; of the sample.First, for each time point recorded, the coordinates of

the sample markers are fit to a quadratic equation usingleast squares regression:

y ¼ ax2 þ bx þ c: ð1Þ

The curvature, k; of the sample at each time point isthen calculated from the relation [16]:

k ¼y00

ð1þ ðy0Þ2Þ3=2: ð2Þ

In which y0 and y00 refer to the first and secondderivatives of the quadratic equation, respectively. Themoment, M ; applied to the sample at each time point iscalculated from the relation:

M ¼ Fxy þ Fyx: ð3Þ

The moment arm x is the horizontal distance from thesecond post to the flexure bar marker. The moment army is the vertical distance from the horizontal linesegment connecting the first and second posts to theflexure bar marker (Fig. 3). Fx and Fy represent thehorizontal and vertical components of the reaction forceat the posts, calculated from the load data and the anglemade between the tangent to the quadratic curve at thesecond post and the horizontal. Based on this data, theMatlab program outputs two plots. The first plot depictsthe marker positions and the fitted quadratic curve forthe reference state and state of maximum curvature(Fig. 5a). The second plot is the graph of moment, M ;versus change in curvature, Dk (Fig. 5b).The Matlab program terminates after calculating the

flexural rigidity of each specimen by conducting anordinary least squares regression on M versus Dk Theregression yields a linear relationship whose slope isthe flexural rigidity of the scaffold sample, as per theBernoulli–Euler moment–curvature relationship [16], inwhich M is the moment applied to the sample, EeffI isthe flexural rigidity, and Dk is the change in curvature:

M ¼ EeffIDk: ð4Þ

Subsequent calculations are performed on a separatespreadsheet in which the second moment of inertia ofthe scaffold sample, I ; is calculated from the followingrelation in which t and w are the thickness and width ofthe sample, respectively:

I ¼1

12t3w: ð5Þ

Finally, the effective stiffness of the scaffold sample,Eeff ; is calculated by dividing the flexural rigidity by I :Physically, Eeff calculated from Eq. 4 is analogous to the

Fig. 4. Representative thickness measurement image from SigmaScan

Pro (SPSS, Inc., Chicago, IL). Lines are traced across the thickness of

the sample at five uniformly spaced locations along its length. The

mean of the five line lengths was used as the thickness in the data

analysis calculations.

G.C. Engelmayr Jr. et al. / Biomaterials 24 (2003) 2523–2532 2527

Young’s Modulus measured in uniaxial tension.Further, since scaffold materials are in general hetero-geneous nonlinear anisotropic materials, Eeff is repre-sentative of the instantaneous effective modulus at aparticular level of Dk and direction of flexure. Thevalues for Eeff determined by these techniquescompare well with standard tensile testing results. Amaximum deviation of three percent was observed forsilicone samples of comparable stiffness (unpublishedobservations).

2.7. Scanning electron microscopy

One scaffold sample representing each flexed andstatic control time point (10 total) was prepared forSEM. Dry scaffold samples were attached to aluminumsample stubs and sputter coated with a thin layer ofgold–palladium alloy (B3 nm thickness) in a Cressington108 auto sputter coater with a Cressington MTM-20thickness controller (Cressington Scientific Instruments,Inc., Cranberry Twp., PA). Each scaffold sample wasimaged at 40, 100, and 300-fold magnification at 5 kV ona JEOL JSM-6330F field emission scanning electronmicroscope (JEOL USA, Inc., Peabody, MA).

2.8. Statistics

Effective stiffness data were expressed as mean7standard error (SEM). Test groups were compared usingunpaired, two-tail student t-tests. Differences betweentest groups were considered statistically significantwhen po0:05:

3. Results

3.1. Bioreactor operation and sterility

The bioreactor demonstrated consistent operationthroughout each of the two runs. Microscopic evalua-tion of the culture medium at each time point indicatedthat sterility was maintained throughout the course ofeach study (up to 5 weeks).

3.2. Effective stiffness

Eeff (mean7SEM) of the virgin (time zero) PGA/P4HB scaffold samples was 844786 kPa (Fig. 6). Eeffvalues were observed to decrease over time for samplesincubated under both flexed and static conditions.A statistically significant decrease (po0:01) was ob-served in Eeff between the virgin sample and the flexedsample at 1 week (�43%); no significant difference wasobserved between the virgin material and the staticcontrol sample at 1 week. Taken together, this indicatesthat dynamic flexure has a significant effect on Eeffcompared to static incubation conditions at 1 week.Substantial loss of stiffness was incurred by both static(�82%) and flexed (�91%) by 3 weeks, with astatistically significant difference between the staticand flexed groups (po0:05).

Change-in-curvature (mm-1)

-0.06 -0.04 -0.02 0.00 0.02 0.04

Mom

ent (

mN

-mm

)

-10

-5

0

5

10 AFWF

X coordinate (microns)

-5000 0 5000 10000 15000 20000

Y c

oord

inat

e (m

icro

ns)

-4000

-3000

-2000

-1000

0

1000

2000

3000

4000

Initial Marker CoordinatesInitial Quadratic FitMax. Marker CoordinatesMax. Quadratic Fit

Post 1 Post 2

(A)

(B)

Fig. 5. Representative marker position (A) and moment–curvature (B)

plots output from the Matlab data analysis program.

Incubation time (weeks)0 1 static 1 flexed 3 static 3 flexed

Eef

f (kP

a)

0

200

400

600

800

1000

1200

N.S.p < 0.01 p < 0.01 p < 0.01

p < 0.05

* * *

*

- 43%

- 82% - 91%

Fig. 6. Effective stiffness of PGA/P4HB samples. Values presented are

the mean of the measurements WF and AF for six samples. Error bars

indicate standard error. Insignificant comparisons are designated by

‘‘N.S.’’ WF and AF designations simply refer to opposing directions

for the virgin (time zero) and static sample groups.

G.C. Engelmayr Jr. et al. / Biomaterials 24 (2003) 2523–25322528

Eeff of the virgin PGA/PLLA/P4HB scaffold sampleswas 16877151 kPa (Fig. 7), approximately twice that ofthe PGA/P4HB. There was no statistically significantdifference in the stiffness of the static control samplesbetween times 0, 3, and 5 weeks. There was, however, astatistically significant decrease in stiffness between theflexed and static control samples at 3 weeks (�72%) and5 weeks (�76%). There was not, however, anysignificant difference in Eeff between the flexed samplesat 3 and 5 weeks.The effective stiffness ratio is calculated here as the

ratio of Eeff measured in the WF direction divided bythat measured in the opposing AF direction. Aneffective stiffness ratio of unity indicates isotropicflexural behavior, whereas any significant deviationfrom unity would suggest anisotropic flexural behavior.The effective stiffness ratio of the PGA/PLLA/P4HBscaffold samples is depicted in Fig. 8. Static controlsamples did not exhibit a statistically significant devia-tion from one, indicating that they were isotropic. Theeffective stiffness ratio of flexed samples, however,exhibited a statistically significant deviation from one(�42%) at 3 weeks and 5 weeks (�29%) (po0:01),indicating that these samples were stiffer when flexed inthe AF direction. The effective stiffness ratio of flexedand static PGA/P4HB samples did not deviate signifi-cantly from one, indicating isotropic behavior (data notshown).

3.3. Scanning electron microscopy

Representative SEM images are depicted for compar-ison (Fig. 9). Several qualitative observations can bemade from the images: (1) the fiber structure andorientation of the virgin scaffold materials appears to be

random within each sample and across samples (datanot shown), (2) the P4HB coating does not appear to beuniform over the fiber structure, but rather appears tolocalize in sparsely distributed aggregates (Figs. 9a andd), (3) fiber diameter does not appear to diminishsignificantly over 3 weeks (PGA/P4HB) or 5 weeks(PGA/PLLA/P4HB), (4) fibers appear to fragment overthe course of the incubation, and this fragmentationappears to be more pronounced in the dynamicallyflexed PGA/P4HB samples (Fig. 9f).

4. Discussion

TEHV represent a promising strategy for the treat-ment of valvular disease, in particular for pediatricapplications in which the growth and remodelingpotential offered by TEHV is necessary for long-termfunction [7]. The ideal TEHV design has yet to beelucidated, as the guiding principles of tissue engineeringare still in flux. In order to facilitate the systematicevaluation of candidate scaffolds and constructs, wedeveloped a bioreactor for the incubation of biomaterialsamples under conditions of cyclic three-point flexure, asimple mode of flexural stimulation. While cyclic three-point flexure is not identical to the mode of dynamicflexure exhibited by native heart valves in vivo, or by aTEHV in a pulse-duplicator, it allows us to isolate theeffect of flexure from other potential modes of stimula-tion, such as pulsatile pressure or shear stress. To ourknowledge, the bioreactor described herein is the firstdevice aimed to study the independent effect of dynamicflexure on TEHV biomaterials. In these first experimentswe have demonstrated the operation of the bioreactorand its ability to maintain sterility for at least 5 weeks in

Incubation time (weeks)

0 3 static 3 flexed 5 static 5 flexed

Eef

f (kP

a)

0

500

1000

1500

2000

2500

N.S. N.S.p < 0.01 p < 0.01

* *

-72% -76%

Fig. 7. Effective stiffness of PGA/PLLA/P4HB samples. Values

presented are the mean of the measurements WF and AF for six

samples. Error bars indicate standard error. Insignificant comparisons

are designated by ‘‘N.S.’’ WF and AF designations simply refer to

opposing directions for the virgin (time zero) and static sample groups.

Incubation time (weeks)

0 3 static 3 flexed 5 static 5 flexed

Eef

f Rat

io (

WF

/AF

)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6N.S.

N.S.

N.S.p < 0.01 p < 0.01

* *

- 42%

- 29%

Fig. 8. Effective stiffness ratio (WF/AF) for PGA/PLLA/P4HB

samples. Values presented are the mean ratios (WF/AF) calculated

from the WF and AF effective stiffness measurements. Error bars

indicate standard error. Insignificant comparisons are designated by

‘‘N.S.’’ WF and AF designations simply refer to opposing directions

for the virgin (time zero) and static sample groups.

G.C. Engelmayr Jr. et al. / Biomaterials 24 (2003) 2523–2532 2529

an incubator environment. In addition, we implementedthe bioreactor to investigate the effect of dynamicflexure on the Eeff of two different TEHV scaffoldmaterials, PGA/P4HB and PGA/PLLA/P4HB. Thesetwo biomaterials were chosen primarily because theyare currently being evaluated as candidate scaffoldsfor the fabrication of TEHV [7]. Based on recentstudies which evaluated the effects of constant fluidflow [17] and dynamic compressive loading [18] on

similar scaffold materials, we hypothesized that dynamicflexure in a non-flowing environment may augment thebulk degradation rate of the polymers or result infatigue of the polymer fibers, with concomitant lossesin effective stiffness. Moreover, because these non-woven fiber meshes scaffolds have been modifiedby P4HB dip-coating, we further hypothesized thatthey may be more susceptible to fatigue (e.g., dis-ruption of dip-coating) under conditions of dynamic

(A)

(E)(B)

(F)(C)

(D)

Fig. 9. Scanning electron micrographs of representative 3-week static PGA/PLLA/P4HB (A, D), 3-week flexed PGA/PLLA/P4HB (B, E), and

3-week flexed PGA/P4HB (C, F) samples at 40-fold (A–C) and 300-fold (D–F) magnification. In particular, note the dramatic difference in thickness

between the dynamically flexed PGA/P4HB sample (C) and the static (A) and flexed (B) PGA/PLLA/P4HB samples at 3 weeks. The higher

magnification images depict the sparse distribution of P4HB (D), as well as the cracking and fragmentation of PGA fibers under dynamic flexure (F).

G.C. Engelmayr Jr. et al. / Biomaterials 24 (2003) 2523–25322530

stimulation, potentially giving rise to unique mechanicalproperties.In accordance with our hypotheses, incubation under

dynamic flexure led to a trend of decreased Eeffcompared to static conditions in both scaffolds, andinduced directional anisotropy (po0:01) in PGA/PLLA/P4HB samples at 3 and 5 weeks. SEM imagesof these samples hinted at some of the structural featuresand changes that may have been responsible for the Eeffresults. The heterogeneity observed in the virgin scaffoldfiber structure, coupled with the observation that P4HBdoes not uniformly coat the samples, helps explain thevariability in Eeff measured for the virgin (time zero)replicates. Fragmentation of fibers, as opposed todiminishing fiber diameter and mass loss, appears tobe a primary mechanism for loss of Eeff in dynamicallyflexed PGA/P4HB samples. It is not clear from thisstudy if dynamic flexure augmented the bulk degrada-tion rate of the polymers, making them more susceptibleto fracture. Interestingly, significant fragmentation offibers was not observed in any of the PGA/PLLA/P4HBscaffolds, indicating that PLLA fibers may shield thePGA fibers from mechanical disruption.Recent studies in our lab have demonstrated that

uncoated PGA/PLLA scaffolds exhibit isotropic flexuralbehavior both at time zero, and after 3 weeks of static ordynamic flexural incubation (unpublished observations).Furthermore, the effective stiffness of these uncoatedPGA/PLLA scaffolds, both at time zero, and after 3weeks of static or dynamic flexural incubation, isapproximately equal to that of the 3 week dynamicallyflexed PGA/PLLA/P4HB scaffolds in the present study.Collectively, these observations indicate that the con-tribution of the P4HB coating to the effective stiffness ofthe PGA/PLLA/P4HB scaffolds is removed by 3 weeksof dynamic flexural incubation. Disruption of the P4HBbonding between the fibers could potentially explain thisbehavior, however this could not be observed from theelectron micrographs. These findings imply that P4HBdip-coating may be useful in modulating the response ofnon-woven PGA/PLLA scaffold flexural properties todynamic flexure. Hypothetically, properties of ananatomical TEHV scaffold might be modified byP4HB dip-coating such that regions of the scaffoldsubjected to dynamic flexure (e.g., the leaflets) becomemore compliant, while regions of the scaffold notsubjected to dynamic flexure (e.g., the aortic root) retaintheir stiffness. Further studies on anatomical TEHVscaffolds incubated in pulse duplicators would benecessary in order to conclusively demonstrate thisbehavior.Non-woven PGA and PLLA scaffolds have been

widely applied in cardiovascular, cartilage, and softtissue engineering [9,19–22]. As the structural integrityof a TEHV scaffold is critical for acute hemodynamicfunction, it is imperative to evaluate the mechanical

properties of candidate TEHV scaffold materials undercontrolled deformation modes resembling those thatmanifest in vivo. Moreover, future advances in tissueengineering will likely stem from an improved under-standing of the biomechanics of tissue growth andremodeling. Thus, the flexure bioreactor can be utilizedto evaluate candidate TEHV scaffolds and constructs, aswell as to investigate the underlying mechanisms bywhich mechanical loading influences tissue development.

Acknowledgements

This research was supported by the following grantsfrom the National Institutes of Health: HL-68-816-01(MSS) and HL-97-005 (JEM). We would like to thankDrs. Simon C. Watkins and Donna Stoltz at theUniversity of Pittsburgh Center for Biologic Imaging(CBI) for use of their scanning electron microscopyfacilities. MSS is an Established Investigator of theAmerican Heart Association.

References

[1] Niklason LE, et al. Functional arteries grown in vitro. Science

1999;284(5413):489–93.

[2] Seliktar D, et al. Dynamic mechanical conditioning of collagen-

gel blood vessel constructs induces remodeling in vitro. Ann

Biomed Eng 2000;28(4):351–62.

[3] Sodian R, et al. New pulsatile bioreactor for fabrication of tissue-

engineered patches. J Biomed Mater Res 2001;58(4):401–5.

[4] Hoerstrup SP, et al. New pulsatile bioreactor for in vitro

formation of tissue engineered heart valves. Tissue Eng

2000;6(1):75–9.

[5] Zeltinger J, et al. Development and characterization of tissue-

engineered aortic valves. Tissue Eng 2001;7(1):9–22.

[6] Weston MW, Yoganathan AP. Biosynthetic activity in heart valve

leaflets in response to in vitro flow environments. Ann Biomed

Eng 2001;29(9):752–63.

[7] Hoerstrup SP, et al. Functional living trileaflet heart valves grown

in vitro. Circulation 2000;102(19 Suppl. 3):III44.

[8] Seliktar D, Nerem RM, Galis ZS. The role of matrix metallo-

proteinase-2 in the remodeling of cell-seeded vascular constructs

subjected to cyclic strain. Ann Biomed Eng 2001;29(11):923–34.

[9] Jockenhoevel S, et al. Cardiovascular tissue engineering: a new

laminar flow chamber for in vitro improvement of mechanical

tissue properties. Asaio J 2002;48(1):8–11.

[10] Mitchell SB, et al. A device to apply user-specified strains to

biomaterials in culture. IEEE Trans Biomed Eng 2001;48(2):

268–73.

[11] Cacou C, et al. A system for monitoring the response of uniaxial

strain on cell seeded collagen gels. Med Eng Phys 2000;22(5):

327–33.

[12] Kim BS, Mooney DJ. Scaffolds for engineering smooth muscle

under cyclic mechanical strain conditions. J Biomech Eng

2000;122(3):210–5.

[13] Vesely I, Boughner D. Analysis of the bending behaviour of

porcine xenograft leaflets and of natural aortic valve material:

bending stiffness, neutral axis and shear measurements. J Biomech

1989;22(6/7):655–71.

G.C. Engelmayr Jr. et al. / Biomaterials 24 (2003) 2523–2532 2531

[14] Iyengar A, et al. Dynamic in-vitro 3D reconstruction of heart

valve leaflets using structured light projection. Ann Biomed Eng

2001;29:963–73.

[15] Gloeckner DC, Billiar KL, Sacks MS. Effects of mechanical

fatigue on the bending properties of the porcine bioprosthetic

heart valve. Asaio J 1999;45(1):59–63.

[16] Frisch-Fay R, Flexible bars. Washington, DC: Butterworths,

1962. p. 220.

[17] Agrawal CM, et al. Effects of fluid flow on the in vitro

degradation kinetics of biodegradable scaffolds for tissue

engineering. Biomaterials 2000;21(23):2443–52.

[18] Thompson DE, Agrawal CM, Athanasiou KA. The effects of

dynamic compressive loading on biodegradable implants of

50–50% polylactic acid-polyglycolic acid. Tissue Eng 1996;

2(1):61–74.

[19] Weng Y, et al. Tissue-engineered composites of bone and cartilage

for mandible condylar reconstruction. J Oral Maxillofac Surg

2001;59(2):185–90.

[20] Brown AN, et al. Combining chondrocytes and smooth muscle

cells to engineer hybrid soft tissue constructs. Tissue Eng

2000;6(4):297–305.

[21] Ameer GA, Mahmood TA, Langer R. A biodegradable composite

scaffold for cell transplantation. J Orthop Res 2002;20(1):16–9.

[22] Gao J, Niklason L, Langer R. Surface hydrolysis of poly(glycolic

acid) meshes increases the seeding density of vascular smooth

muscle cells. J Biomed Mater Res 1998;42(3):417–24.

G.C. Engelmayr Jr. et al. / Biomaterials 24 (2003) 2523–25322532