Biorheology 43 (2006) 215222 IOS Press

215

The effect of hydrodynamic shear on 3D engineered chondrocyte systems subject to direct perfusion
Manuela T. Raimondi a, , Matteo Moretti a , Margherita Ciof a , Carmen Giordano b , Federica Boschetti a , Katia Lagan a and Riccardo Pietrabissa a
Laboratory of Biological Structure Mechanics, Department of Structural Engineering, Politecnico di Milano, Milano, Italy b Laboratory of Biocompatibility and Cell Cultures, Department of Chemistry, Materials and Chemical Engineering Giulio Natta, Politecnico di Milano, Milano, Italy
Abstract. Bioreactors allowing direct-perfusion of culture medium through tissue-engineered constructs may overcome diffusion limitations associated with static culturing, and may provide ow-mediated mechanical stimuli. The hydrodynamic stress imposed on cells within scaffolds is directly dependent on scaffold microstructure and on bioreactor conguration. Aim of this study is to investigate optimal shear stress ranges and to quantitatively predict the levels of hydrodynamic shear imposed to cells during the experiments. Bovine articular chondrocytes were seeded on polyestherurethane foams and cultured for 2 weeks in a direct perfusion bioreactor designed to impose 4 different values of shear level at a single ow rate (0.5 ml/min). Computational uid dynamics (CFD) simulations were carried out on reconstructions of the scaffold obtained from micro-computed tomography images. Biochemistry analyses for DNA and sGAG were performed, along with electron microscopy. The hydrodynamic shear induced on cells within constructs, as estimated by CFD simulations, ranged from 4.6 to 56 mPa. This 12-fold increase in the level of applied shear stress determined a 1.7-fold increase in the mean content in DNA and a 2.9-fold increase in the mean content in sGAG. In contrast, the mean sGAG/DNA ratio showed a tendency to decrease for increasing shear levels. Our results suggest that the optimal condition to favour sGAG synthesis in engineered constructs, at least at the beginning of culture, is direct perfusion at the lowest level of hydrodynamic shear. In conclusion, the presented results represent a rst attempt to quantitatively correlate the imposed hydrodynamic shear level and the invoked biosynthetic response in 3D engineered chondrocyte systems. Keywords: Tissue engineering, cartilage, mechanobiology, biosynthesis, computational uid dynamics, simulation, porous biomaterials
a

1. Introduction To date, tissue-engineered constructs are the rst step towards the restoring of injured cartilage by cellular approaches. Mass transfer is a critical issue when culturing cells in porous scaffolds, to ensure that adequate nutrient and pH levels are maintained and to eliminate catabolites from the construct. Ex vivo tissue engineering processes involving articular cartilage are particularly dependent on mechanical stimuli for differentiation [9]. Direct perfusion bioreactors have been developed to culture cell/scaffold constructs under improved mass transfer and under hydrodynamic loading [3,5,8,10]. In such systems, the
* Address for correspondence: Manuela Teresa Raimondi, PhD, Dipartimento di Ingegneria Strutturale, Politecnico Di Milano, Piazza Leonardo da Vinci, 32, 20133 Milano, Italy. Tel.: +39 02 66214939; Fax: +39 02 66214939; E-mail: manuela. raimondi@polimi.it.

0006-355X/06/$17.00 2006 IOS Press and the authors. All rights reserved

216

M.T. Raimondi et al. / The effect of hydrodynamic shear on 3D engineered chondrocyte systems

entire construct is directly perfused with culture medium, and the transport of oxygen and nutrients from the medium to cells occurs via both diffusion and convection. Increasing the ow rate increases the shear stresses exerted on cells. This complexity presents serious hurdles in determining adequate ow rates and medium solute concentrations for maintaining cell viability and modulating matrix synthesis [2,7]. Flow determines a range of local uid velocities in the scaffold, which themselves are impossible to measure. Most available models of ow through porous structures are homogeneous models, which implicitly average out microscopic details of the porous geometry to arrive at average permeability, velocity, shear stress, etc. These models have important limitations in situations involving spatially inhomogeneous micro architectures. Therefore, more detailed pore-scale computer simulations of uid transport in tissue-engineering scaffolds populated with living cells have been recently proposed [11,12]. In this work, we rst established a direct-perfusion bioreactor to engineer cartilaginous tissue starting from primary bovine chondrocytes seeded into a 3D polymer scaffold. The bioreactor is conceived in order to impose varying levels of hydrodynamic loading to constructs as the only culture variable. We, then, developed a 3D computational model to quantitatively predict the levels of hydrodynamic shear imposed to cells during the experiments. We used this system as a model to establish a quantitative relationship between the applied shear level and the stage of development of constructs, as dened by the sulfated glycosaminoglycan (sGAG) and DNA content, and by their ratio. 2. Materials and methods 2.1. Direct-perfusion bioreactor In a direct perfusion system, the culture medium is conveyed through the porous structure of cellular constructs, thus exposing the cells located in the entire scaffold thickness both to convective solute transport and to a ow-induced mechanical stimulus. Our system is composed of 4 culture chambers, each carrying constructs shaped as discs, 1 mm in thickness, with a perfused section of 2, 3, 4 or 7 mm in diameter. Each chamber carries 3 constructs of identical size (Fig. 1). The bioreactor consists of two independent circuits. Each circuit (Fig. 2) consists of two chambers connected in series by gas permeable silicone tubing and of a de-bubbler placed between them, along with a medium reservoir. The priming volume in each circuit is 20 ml. The two circuits share a peristaltic pump (Watson-Marlow Bredel Inc., Wilmington, MA, USA). The pump is remotely controlled by a timer which automatically reverses the direction of ow at constant time intervals, to avoid cell migration and to guarantee a homogeneous nutrient distribution on both sides. At a constant value of ow rate imposed by the pump, the average uid velocity induced within the constructs will be different for each of the four different diameter scaffold groups, i.e. higher in smaller constructs and lower in bigger constructs, depending on their section area. This experimental set-up allows to apply four different levels of hydrodynamic stimulus as the only culture variable between the four chambers. 2.2. Cell culture and assays Articular cartilage was harvested aseptically from the metacarpophalangeal joints of 8-month old calves slaughtered at the abattoir. Chondrocyte isolation was carried out as described elsewhere [4]. The reagents used were from Sigma-Aldrich S.r.l. (Italy). The complete medium contained Dulbeccos Modied Eagle Medium (DMEM), sodium pyruvate, Hepes buffer, antibiotics (penicillin/streptomycin),

M.T. Raimondi et al. / The effect of hydrodynamic shear on 3D engineered chondrocyte systems

217

(a)

(b)

Fig. 1. Bioreactor chamber. (a) The chamber contains three scaffolds by means of a scaffold holder enclosed between shells which are sealed by simple pressure. (b) The chamber mounted inside the circuit.

Fig. 2. Scheme of half of the hydraulic circuit. The complete bioreactor consists of two circuits identical to the one represented, mounted in parallel.

L-glutamine, L-ascorbic acid, insulin and a 10% aliquot of foetal calf serum (FCS). Viable cells were counted with trypan blue exclusion and used for scaffold seeding. The scaffold used is DegraPol R , a biodegradable polyestherurethane foam developed at the Swiss Federal Institute of Technology (ETH) in Zurich [14]. The scaffolds employed in our study were sterile discs, 1 mm in thickness and 3, 4, 5 and 8 mm in diameter. The discs were positioned in a culture cluster and seeded with cells at a density of about 80,000 cells/mm3 . The discs were incubated for 24 hours at 37 C to allow for cell adhesion. The seeded constructs were, then, either placed in the bioreactor to

218

M.T. Raimondi et al. / The effect of hydrodynamic shear on 3D engineered chondrocyte systems

be subjected to direct perfusion or in culture plates as static controls. The ow rate was regulated at 0.5 ml/min, the ow direction was cyclically inverted every 40 minutes. Constructs were incubated for 2 weeks, with medium completely replaced every 5 days. Three independent experiments were carried out. At 2 weeks of culture, constructs were prepared for biochemical assays and for electron microscopy. Portions used for biochemical analysis were digested with papain and assayed for DNA content by Hoechst 33258 dye binding [13] and for sGAG content by 1.9-dimethylmethylene blue chloride [6]. Other portions of the constructs were prepared for scanning electron microscopy as previously described [12] and observed at 10 kV using a Stereoscan S260 electron microscope (Leica Cambridge Ltd., Cambridge, UK). 2.3. Computational model This part of the study allowed to characterise the local uid dynamics to which cells were exposed within the perfused constructs. We developed a simulation, at the pore scale, of culture medium ow through the scaffold used in the experiments, using computational uid dynamics (CFD). The calculations and theory below are described in detail elsewhere [1]. Briey, a 3D solid model of the scaffold micro-geometry was reconstructed from 250 micro-computed tomography (CT) images (Fig. 3). This scaffold is characterised by an interconnected porosity of roughly 77% and nominal pore size of 100 m. Cell dimension was assumed to be negligible if compared to pore size. The uid domain was created by meshing the void volume of the 3D solid model. The commercial nite-volume code Fluent (Fluent Inc., Lebanon, NH, USA) was used to set up and solve the CFD problem. In order to quantify the uid-dynamic environment imposed to cells in each culture chamber, simulations were run by imposing the calculated inlet velocities as boundary conditions. The inlet velocity specic to each scaffold size was calculated from the ow rate adopted in the experiments (0.5 ml/min for each bioreactor chamber

Fig. 3. Method employed to set up the CFD model. (a) SEM image of the scaffold microstructure, (b) micro-computed tomography (CT) image acquired on the scaffold, (c) mesh of the uid domain reconstructed for a cubic scaffold portion, 400 m in side.

M.T. Raimondi et al. / The effect of hydrodynamic shear on 3D engineered chondrocyte systems

219

holding three scaffolds). The uid-induced shear stresses acting at the wall of the pores were calculated and assumed as an estimate of the shear stresses acting on the membranes of the cells. The mean, mode and median shear stress values were determined.

3. Results The results of the uid dynamic simulations were analyzed at the central walls of the reconstructed uid domain, to avoid boundary effects. The average, median and mode shear stress values calculated at the scaffold walls are shown in Table 1 for each size of perfused construct. Both the inlet uid velocity and the median shear stresses showed a 12-fold increase between the 7 mm and the 2 mm constructs, consistently with the 12-fold reduction in perfused section area. The predicted median shear within constructs ranged from 4.6 mPa (0.046 dyn/cm2 ) for the 7 mm constructs to 56 mPa (0.56 dyn/cm2 ) for the 2 mm constructs. Figure 4 shows the results of the biochemical assays at 2 weeks of culture, in terms of content in DNA and sGAG specic to construct volume, and in terms of sGAG/DNA. Values are plotted for increasing values of median shear stress and are given as the mean and standard deviation of 3 independent experiments. The mean content in DNA and in sGAG increased at increasing levels of applied shear stress. The total increase, correspondent to the 12-fold increase in the level of applied shear, was in the order of 1.7-fold and 2.9-fold for DNA and sGAG content, respectively. In contrast, the mean sGAG/DNA ratio, indicative of the biosynthesis of sGAG specic to cells, showed a tendency to decrease for increasing shear levels, although the standard deviation calculated for this parameter was quite high. The content in DNA and sGAG estimated for the non-perfused controls was comparable to that estimated for constructs subjected to a moderate regimen of hydrodynamic stimulus (i.e. a shear stress level of 25 mPa). At the SEM, chondrocytes showed a phenotypic spherical morphology with smooth membranes in all culture conditions (Fig. 5). Cells were dispersed within large aggregates of neo-formed matrix. Surface morphology of the synthesised matrix was rougher in constructs non-perfused and in constructs cultured at lower shear level (Fig. 5a), and smoother in constructs cultured at higher shear levels (Figs 5b,c).

4. Discussion The bioreactor for direct perfusion presented in this study was designed with the aim to establish a quantitative relationship between the level of hydrodynamic shear applied to 3D engineered chondrocyte systems and parameters related to the invoked biosynthetic response. Our main nding was that
Table 1 Results of the CFD simulations Diameter of perfused construct [mm] 2 3 4 7 Experimental ow rate [ml/min] 0.5 0.5 0.5 0.5 Construct inlet velocity [m/sec] 884 393 221 72 Mean shear stress [mPa] 63 28 16 5.2 Median shear stress [mPa] 56 25 14 4.6 Mode shear stress [mPa] 38 17 9.4 3.1

220

M.T. Raimondi et al. / The effect of hydrodynamic shear on 3D engineered chondrocyte systems

(a)

(b)

(c)
Fig. 4. Results of the biochemical assays for the chondrocyte-seeded constructs at 2 weeks of direct-perfusion under varying hydrodynamic conditions. (a) DNA content, (b) sGAG content and (c) sGAG/DNA content. Values are reported for increasing values of median shear stress (zero stress refers to non-perfused controls) and are given as mean standard deviation of 3 independent experiments.

at increasing levels of hydrodynamic loading, in the range 5.6 to 56 mPa, sGAG and DNA content increased after two weeks of culture, whereas sGAG specic to cells showed a tendency to decrease. Few comparisons with literature data exist. Our results are in agreement with those obtained by Saini and Wick [15], both in terms of sGAG and sGAG/cell content as a function of increasing levels of hydrodynamic loading. In the mentioned work, hydrodynamic shear was imposed at higher levels and only at the construct surface, thus only qualitative agreement may be expected. In most previous studies of direct-perfusion [5,8,10], constructs were cultured under a xed uid velocity and the level of shear induced on cells was never quantied. These studies have demonstrated an increase in cell content and matrix synthesis in perfused constructs, as compared to static controls, again in general agreement with our results. Only one study investigated the effect of direct perfusion at two levels of velocity [3]. Here, a higher uid velocity was found to decrease cell content, as compared to a lower uid velocity, when applied at early time points in culture. As a matter of fact, matrix content in cellular constructs is minimal at the early period of culture. It is possible that perfusion enables uid shear to wash out cells and matrix from the scaffold, consistently with smoothening of the matrix surface observed in our study. This hypothesis could explain both the ndings by Davisson [3], and our nding that content in DNA and sGAG estimated for the non-perfused controls was higher with respect to constructs subject to the lowest regimen of hydrodynamic loading (i.e. 4.6 mPa). However, cell and sGAG content in the medium were not measured so this hypothesis remains to be veried. The bioreactor system was able to induce signicantly different uid dynamic environments within constructs cultured in the four different chambers. The magnitude of medium velocity imposed at the

M.T. Raimondi et al. / The effect of hydrodynamic shear on 3D engineered chondrocyte systems

221

(a)

(b)

(c)
Fig. 5. SEM images of chondrocyte-seeded constructs at 2 weeks of culture under direct-perfusion. Surface morphology of the synthesised matrix varies in constructs subject to increasing levels of median hydrodynamic shear of: (a) 4.6 mPa, (b) 14 mPa and (c) 56 mPa.

construct inlet ranged from around 70 m/sec in 7 mm constructs to around 900 m/sec in 2 mm constructs. The level of hydrodynamic shear induced on cells within constructs, as estimated by CFD simulation, ranged from 4.6 mPa in 7 mm constructs to 56 mPa in 2 mm constructs. This level of shear is of the same order of magnitude as the level to which chondrocytes are exposed within natural articular cartilage (100 mPa for a 10 MPa joint contact pressure, as estimated by Schinagl et al. [16]). The magnitude of shear stresses in this study was calculated from a numerical analysis of ow, which is yet to be validated. Unfortunately, local hydrodynamic shear cannot be measured within constructs. Our predicted levels of shear stress agree with other theoretical predictions [11] referred to a different 3D scaffold. Again, only a general agreement may be expected with literature data in this respect, as the local shear stresses experienced by cells can be signicantly different, even for the same inlet uid velocity, as a function of the scaffold microarchitecture. In conclusion, the results presented here represent the rst attempt to quantitatively correlate the imposed level of hydrodynamic shear and the invoked biosynthetic response in 3D engineered chondrocyte systems. Our results suggest that the optimal condition to favour sGAG synthesis in engineered constructs, at least at the beginning of culture, is direct perfusion at the lowest level of hydrodynamic shear technically achievable.

222

M.T. Raimondi et al. / The effect of hydrodynamic shear on 3D engineered chondrocyte systems

Acknowledgements This project is funded by Fondazione Cariplo (grant #2004.1148) and by Politecnico di Milano (grant CelTec). The DegraPol scaffolds were provided without charge by Dr. Peter Neuenschwander of ETH, Zurich. References
[1] M. Ciof, F. Boschetti, M.T. Raimondi and G. Dubini, Modeling evaluation of the uid-dynamic microenvironment in tissue-engineered constructs: A micro-CT based model, Biotechnol. Bioeng. 93(3) (2006), 500510. [2] E.M. Darling and K.A. Athanasiou, Articular cartilage bioreactors and bioprocesses, Tissue Eng. 9(1) (2003), 926. [3] T. Davisson, R.L. Sah and A. Ratcliffe, Perfusion increases cell content and matrix synthesis in chondrocyte threedimensional cultures, Tissue Eng. 8(5) (2002), 807816. [4] O. Demarteau, D. Wendt, A. Braccini, M. Jakob, D. Schafer, M. Heberer and I. Martin, Dynamic compression of cartilage constructs engineered from expanded human articular chondrocytes, Biochem. Biophys. Res. Commun. 310(2) (2003), 580588. [5] N.S. Dunkelman, M.P. Zimber, R.G. LeBaron, R. Pavelec, M. Kwan and A.F. Purchio, Cartilage production by rabbit articular chondrocytes on polyglycolic acid scaffolds in a closed bioreactor system, Biotechnol. Bioeng. 46 (1995), 299 305. [6] R.W. Farndale, C.A. Sayers and A.J. Barrett, A direct spectrophotometric microassay for sulphated glycosaminoglycans in cartilage cultures, Connective Tissue Research 9 (1982), 247248. [7] I. Martin, D. Wendt and M. Heberer, The role of bioreactors in tissue engineering, Trends Biotechnol. 22(2) (2004), 8086. [8] S. Mizuno, F. Allemann and J. Glowacki, Effects of medium perfusion on matrix production by bovine chondrocytes in three-dimensional collagen sponges, J. Biomed. Mater. Res. 56(3) (2001), 368375. [9] V.C. Mow and C.C. Wang, Some bioengineering considerations for tissue engineering of articular cartilage, Clin. Orthop. 367 (1999), S204S223. [10] D. Pazzano, K.A. Mercier, J.M. Moran, S.S. Fong, D.D. DiBiasio, J.X. Rulfs, S.S. Kohles and L.J. Bonassar, Comparison of chondrogensis in static and perfused bioreactor culture, Biotechnol. Prog. 16(5) (2000), 893896. [11] B. Porter, R. Zauel, H. Stockman, R. Guldberg and D. Fyhrie, 3-D computational modeling of media ow through scaffolds in a perfusion bioreactor, J. Biomech. 38(3) (2005), 543549. [12] M.T. Raimondi, F. Boschetti, L. Falcone, G.B. Fiore, A. Remuzzi, M. Marazzi, E. Marinoni and R. Pietrabissa, Mechanobiology of engineered cartilage cultured under a quantied uid dynamic environment, Biomech. Modell. Mechanobiol. 1 (2002), 6982. [13] J. Rao and W.R. Otto, Fluorimetric assay for cell growth estimation, Anal. Biochem. 207 (1992), 186192. [14] B. Saad, Y. Kuboki, M. Welti, G.K. Uhlschmid, P. Neuenschwander and U.W. Suter, DegraPol-foam: a degradable and highly porous polyesterurethane foam as a new substrate for bone formation, Artif. Organs 24(12) (2000), 939945. [15] S. Saini and T.M. Wick, Concentric cylinder bioreactor for production of tissue engineered cartilage: effect of seeding density and hydrodynamic loading on construct development, Biotechnol. Prog. 19(2) (2003), 510521. [16] R.M. Schinagl, M.S. Kurtis, K.D. Ellis, S. Chien and R.L. Sah, Effect of seeding duration on the strength of chondrocyte adhesion to articular cartilage, J. Orthop. Res. 17(1) (1999), 121129.