The contribution of matrix and cells to leaflet retraction in heart valve tissue engineering

Transcription

1 The contribution of matrix and cells to leaflet retraction in heart valve tissue engineering

2 A catalogue record is available from the Eindhoven University of Technology Library ISBN: Copyright 2011 by M.A.A. van Vlimmeren All rights reserved. No part of this book may be reproduced, stored in a database or retrieval system, or published, in any form or in any way, electronically, mechanically, by print, photo print, microfilm or any other means without prior written permission by the author. Cover design: Ineke van Vlimmeren Printed by Ipskamp Drukkers B.V., Enschede, the Netherlands. Financial support by the Dutch Heart Foundation for the publication of this thesis is gratefully acknowledged. The authors gratefully acknowledge the support of the Smart Mix Program of the Netherlands Ministry of Economic Affairs and the Netherlands Ministry of Education, Culture and Science.

3 The contribution of matrix and cells to leaflet retraction in heart valve tissue engineering PROEFSCHRIFT ter verkrijging van de graad van doctor aan de Technische Universiteit Eindhoven, op gezag van de rector magnificus, prof.dr.ir. C.J. van Duijn, voor een commissie aangewezen door het College voor Promoties in het openbaar te verdedigen op donderdag 3 november 2011 om uur door Marijke Antonia Adriana van Vlimmeren geboren te Veldhoven

4 Dit proefschrift is goedgekeurd door de promotor: prof.dr.ir. F.P.T. Baaijens Copromotoren: dr. A. Driessen Mol en dr.ir. C.W.J. Oomens

5 Contents Summary III Chapter 1: General introduction The human heart valves Heart valve disease and replacements Heart valve tissue engineering Tissue stress, compaction and retraction Collagen maturation Rationale and outline 14 Chapter 2: An in vitro model system to quantify stress generation, compaction and retraction in engineered heart valve tissues Introduction 2.2 Materials & Methods 2.3 Results 2.4 Discussion 2.5 Acknowledgements Chapter 3: Passive and active contributions to generated force and retraction in heart valve tissue engineering Introduction 3.2 Materials & Methods 3.3 Results 3.4 Discussion 3.5 Acknowledgements Chapter 4: Controlling matrix formation and cross linking by hypoxia in cardiovascular tissue engineering Introduction 4.2 Materials & Methods 4.3 Results 4.4 Discussion 4.5 Acknowledgements I

6 Contents Chapter 5: Low oxygen concentrations impair tissue development in tissue engineered cardiovascular constructs Introduction 5.2 Materials & Methods 5.3 Results 5.4 Discussion 5.5 Acknowledgements Chapter 6: The potential of prolonged tissue culture to reduce stress generation and retraction in engineered heart valve tissues Introduction 6.2 Materials & Methods 6.3 Results 6.4 Discussion 6.5 Acknowledgements Chapter 7: General discussion Introduction and main findings Limitations and considerations Implications and future perspectives Conclusion 114 References 115 Samenvatting 131 Dankwoord 133 Curriculum vitae 135 List of publications 137 II

7 Summary The contribution of matrix and cells to leaflet retraction in heart valve tissue engineering Heart valve tissue engineering is a promising technique to overcome the drawbacks of currently used mechanical and prosthetic heart valve replacements. Tissue engineered (TE) heart valves are viable and autologous implants that have the capacity to grow, remodel and repair throughout a patient s life, without the need of anticoagulation therapy. The valves are made by seeding extracellular matrix (ECM) producing cells, such as vascularderived cells, onto a rapidly degrading scaffold material manufactured into the shape of a heart valve. TE valves are cultured constraint whereby the leaflets fuse together. During four weeks of culture, the scaffold is replaced by newly formed tissue, while stress is generated within the tissue by traction forces exerted by the cells. This stress is beneficial for tissue formation and architecture. However, during culture it causes tissue compaction, resulting in leaflet flattening, and at time of implantation, the leaflet constraints are released and the generated stress causes retraction of the leaflets. Due to this retraction, the leaflets are not able to fully close during diastole and valvular regurgitation occurs. In this thesis, this phenomenon of tissue retraction was examined and strategies to decrease retraction were investigated. The first aim was to unravel and quantify stress, compaction and retraction in developing engineered heart valve tissues. Therefore, a representative in vitro model system of rectangular TE strips (TE constructs) was developed in which the evolution of stress and compaction during culture, and the resulting retraction after release of constraints, can be quantified from a single TE construct. An important finding was that during the first 2 weeks of tissue culture, the scaffold was able to counterbalance the traction forces of the cells, which reveals a key role of the stiffness of the cellular surroundings in compaction. When scaffold degradation started after 2 weeks, stress generation and compaction became evident and gradually continued up to week 4. In average, the TE constructs compacted 50 65% in width and reached force levels of mn and stress levels of 5 30 kpa. The resulting retraction 24 hours after release of constraints was 35 50% in length. These degrees of compaction and retraction thus seriously affect leaflet geometry and need substantial reduction to prevent regurgitation. Subsequently, the relative contributions of passive and active retraction were examined. Passive retraction occurs through passive stress release in the cells and ECM at release of constraints, while active retraction is caused by the traction forces of the cells. To quantify the active and passive contributions, the active traction forces of the cells were eliminated by Cytochalasin D or an inhibitor of the Rho associated kinase pathway, while both passive and active contributions of the cells were eliminated by lysis and/or removal of III

8 Summary the cells. A major finding in this study was that the passive contribution of cells to retraction is substantial. It was found that passive cell retraction accounts for 45% of total retraction, while active cell retraction accounts for 40% of the total retraction. The remaining 15% is attributed to passive retraction of the ECM. These findings illustrate the importance of the cells in the process of tissue retraction, not only actively retracting the tissue, but also in a passive manner. Finally, we aimed to decrease tissue retraction in order to obtain functional, nonretracting leaflets. As it was hypothesized that a strong and well developed ECM would provide more resistance to the cell traction forces, two strategies to improve the mechanical integrity of the ECM were investigated. First, the effect of the environmental factor oxygen concentration on the ECM formation was investigated at both cell (2D) and tissue (3D) level. At cellular level, culturing at oxygen concentrations of 4% and below (hypoxia) enhanced the production of collagen cross link enzymes and to a lesser extent collagen type I and III. Unfortunately, these results did not translate into enhanced collagen deposition and maturation in TE constructs. Tissue properties remained similar at 7% and 4% O 2 as compared to 21% O 2, while culturing below 4% O 2 reduced ECM production and the mechanical integrity of the tissue. From this latter study, it was concluded that hypoxia is not very likely to create a more robust ECM. In the second strategy, the ECM integrity was hypothesized to increase by prolonged tissue culture. Collagen content and cross linking remained constant when increasing culture time from 4 weeks to 6 and 8 weeks, but GAG content increased, which resulted in thicker tissues. Although the generated force remained constant from week 4 on, the increased thickness contributed to a decrease in generated stress. The most important finding in this study was that retraction decreased by ~50% at week 6 and 8 compared to week 4, likely due to the increased GAG content. These findings emphasize the role of the ECM in tissue retraction and that changing its composition might represent an important strategy to reduce tissue retraction and, thereby, valvular regurgitation. To summarize, solving the problem of leaflet retraction in heart valve tissue engineering remains a challenge due to the substantial contribution of passive retraction. The cellular surroundings have shown to affect the resulting compaction and retraction. However, improving resistance against cell traction forces by enhancing the compressive stiffness of the ECM has proven to be difficult. Therefore, other promising approaches like a slower degrading scaffold, adjustment of the valve geometry or decellularization of the tissue are discussed in the present thesis and will be focus of future research. Although these strategies still require extensive research, the new insights into the mechanisms of leaflet retraction obtained within this thesis provide useful knowledge needed to deal with the problem in order to develop functional tissue engineered heart valves. IV

9 Chapter 1 General introduction

10 Chapter 1 Heart valve tissue engineering is a promising approach to overcome the limitations of current available heart valve replacements. However, there are some challenges that need to be overcome, before safe transition to clinical use is possible. The present thesis will unravel the problem of cell traction mediated tissue shrinkage in heart valve tissue engineering. A model system is developed to investigate and quantify the aspects of tissue shrinkage and problem solving strategies are investigated. Before going into details, first some background information about native and tissue engineered heart valves is provided in the next section. 1.1 The human heart valves The human heart is an essential muscular organ that regulates the blood flow through the body. Blood is oxygenated in the lungs through the pulmonary circulation and subsequently pumped through the systemic circulation where it delivers oxygen to the organs, tissues and cells and removes carbon dioxide. Unidirectional blood flow within the heart is regulated by four heart valves (Figure 1.1). The atrio ventricular tricuspid and mitral valves prohibit reverse flow from the ventricles to the atria during systole. The pulmonary and aortic semilunar valves prevent retrograde flow back from the pulmonary and aortic arteries into the ventricles during diastole. Heart valves are complex biological tissues capable of sustaining the successive loading of approximately 40 million heart cycles a year, equivalent to approximately 3 billion cycles over a typical 75 year lifetime (Schoen, 2011). Figure 1.1: Schematics of the human heart and its four heart valves. (A) Cross section of the heart, anterior view, with the tricuspid valve positioned between the right atrium and right ventricle and the mitral valve positioned between the left atrium and left ventricle. In addition, the pulmonary valve is situated between the outlet of the right ventricle and the pulmonary artery, and the aortic valve is situated between the outlet of the left ventricle and the aorta. (B) Cross section of the heart, top view, showing the four heart valves from above. 2

11 General Introduction Basic anatomy and physiology of the semilunar valves The semilunar valves are positioned between the outlet of the right ventricle and the lungs (pulmonary valve) and the outlet of the left ventricle and the aorta (aortic valve). They are referred to as semilunar valves due to the half moon shape of their three flexible leaflets (Figure 1.2). The leaflets are attached to a ring of tough fibrous tissue, called the annulus. The attachment points of two adjacent leaflets to the root are named the commissures. Behind each leaflet are dilated pockets in the vessel wall; the sinuses of valsalva. In the middle of the free edge of each leaflet is a fibrous section called the nodule of Arantius. Coaptation of the three nodules ensures complete central closure of the valve. Figure 1.2: Schematic anatomy of one leaflet of the aortic valve. (A) Cross sectional view indicating the leaflets attached to the annulus fibrosus with the sinus of valsalva behind the leaflet. In addition, the coronary orifice in the aortic wall is shown. (B) Front view of one leaflet indicating the commissures, nodule of arantius and the annulus fibrosus. (Adapted from cardiac surgery in the adult, L.H. Cohn) There are several differences between the aortic and the pulmonary valves, including the size and structure of their leaflets and the fact that the aortic valve has coronary artery orifices behind two of its leaflets that provide blood to the heart muscle. Furthermore, the mechanical loading of the valves during the cardiac cycle is different. The cardiac cycle consists of a period of relaxation called diastole, during which the heart valves are closed and the heart fills with blood, followed by a period of contraction called systole, during which the valves are pushed open. The pressure difference over the valve leaflets during the diastolic phase is significantly higher in the aortic valve than in the pulmonary valve. In the adult heart, transvalvular pressure at pulmonary position is 10 mmhg just before opening of the valves, while transvalvular pressure over the aortic valve reaches 80 mmhg (Guyton and Hall, 2000). 3

12 Chapter The heart valve leaflets The mechanical functioning of the semilunar heart valves depends on the elasticity and structural integrity of the thin leaflets. The essential functional components of the leaflets comprise cells and extracellular matrix (ECM). Two cell types exist, namely, the valvular endothelial cells (VECs) and the valvular interstitial cells (VICs). The VECs form a single layer of cells lining the leaflet surface, providing a protective non thrombogenic surface. VICs are the most abundant cell type in leaflets and are distributed throughout all layers. In healthy adult heart valves, VICs have a quiescent fibroblast like phenotype, but they can differentiate into myofibroblasts to mediate ECM remodeling and valve repair (Rabkin Aikawa et al., 2004; Mendelson and Schoen, 2006; Schoen, 2008; Apte et al., 2011). In developing heart valves, VICs have an activated myofibroblast phenotype as well to mediate growth (Rabkin Aikawa et al., 2004). The ECM of the leaflets consists of collagen, elastin, and amorphous ECM, composed predominately of glycosaminoglycans (GAGs). The leaflets have a layered architecture, in which three layers can be distinguished (Figure 1.3A); the ventricularis, the spongiosa and the fibrosa, which have different compositions and mechanical properties (Flanagan and Pandit, 2003; Schoen, 2008; Apte et al., 2011; Bouten et al., 2011). The fibrosa, at the outflow side of the leaflet, provides primary strength to the leaflet consisting mainly of collagen fibers. At the commissures, the collagen fibers are predominantly oriented in the circumferential direction, while more diverging fibers are present in the centre (Figure 1.3B). Figure 1.3: (A) Schematic cross section of the aortic leaflet indicating the fibrosa, spongiosa and ventricularis (Vesely, 1998). (B) Anisotropic collagen architecture within the aortic valve. Collagen fibers are oriented in circumferential direction at the commissures and more divergent in the center of the leaflet (adapted from Balguid et al., 2007). 4

13 General Introduction This natural fiber orientation results in anisotropic behavior of the leaflet with high flexibility in the radial direction and high strength in the circumferential direction (Balguid et al., 2007). The spongiosa forms the central core of the leaflets and predominantly consists of GAGs. GAGs have a significant water binding capacity, due to their highly negative charge (Culav et al., 1999). This GAG water combination of the spongiosa absorbs shear and provides compressive strength to the ventricularis and fibrosis. By contrast, the ventricularis is an elastin containing layer that ensures flexibility of the leaflets and recoil during systole (Vesely, 1998). 1.2 Heart valve disease and replacements Pathological changes associated with heart valves consist largely of four types: (1) divergent valve architecture as in congenital malformation, (2) ECM damage leading to loss of mechanical integrity of the leaflets (e.g. myxomatous disease, infective endocarditis), (3) nodular calcification, (4) fibrotic thickening (e.g. due to rheumatic fever). These disorders can lead to heart valve insufficiency or stenosis. In insufficient heart valves leakage occurs, causing backflow from the artery to the ventricle during diastole (regurgitation). In case of stenosis, the valve opening becomes narrowed and the heart has to increase its contraction force to pump sufficient amounts of blood through the affected valve. Valvular heart disease is a significant cause of morbidity and mortality worldwide. Although heart valve repair is the preferred treatment, this is not always possible. Worldwide, approximately heart valve replacements are performed annually and due to an ever aging world population, this number has been estimated to increase threefold over the upcoming five decades (Yacoub and Takkenberg, 2005; Pibarot and Dumesnil, 2009). Current available heart valve replacements are either mechanical or bioprosthetic in nature. The first available mechanical valve was the ball and cage valve in Since then, many modifications have been made to this valve to improve its performance. Current mechanical valves are made of carbon and the bi leaflet mechanical valve is, at present, the most popular design. Mechanical heart valves are long lasting and readily available, but vulnerable to thrombus formation, due to high shear stresses and nonphysiological flow profiles that result in blood damage (Dasi et al., 2009). As a consequence, lifelong anticoagulation treatment is required, which involves increased risk of internal bleeding. Bioprosthetic valves are either of human (homograft) or animal origin (xenograft). The homograft valves are intact cryopreserved human valves obtained from donors and are closest to natural valves, but their use is restrained due to limited availability of donors and high costs. Xenografts are made of gluteraldehyde fixated porcine or bovine material. The major advantage of bioprosthetic 5

14 Chapter 1 valves is that there is no need for anticoagulation therapy and that the mechanical functioning of these valves is very good as they have a native like shape. However, these valves are prone to structural degeneration and calcification (Pibarot and Dumesnil, 2009; Siddiqui et al., 2009), and the associated need for reoperations makes them less suitable for young patients. Furthermore, the use of xenografts is associated with the risk of zoonoses, which are human diseases caused by infectious agents from animals. The overall limitation of the currently available heart valve replacements is that they are non viable prostheses, unable to adapt to the constantly alternating hemodynamic environment. In essence, these valves lack the capability of growth, repair and remodeling in the body. These shortcomings restrict the use of these valves in pediatric patients as multiple operations are required to accommodate growth during their childhood. Although both mechanical and bioprosthetic heart valve prostheses significantly improve life expectancy, the ideal prosthetic heart valve, able to adapt to its environment, has yet to be developed. 1.3 Heart valve tissue engineering Tissue engineering aims to develop autologous heart valves able to function a lifetime in vivo. Several key characteristics required for a tissue engineered heart valve are viability, sufficient strength to withstand repetitive and substantial mechanical stress, and the ability to grow and repair any damage by tissue remodeling. Several tissue engineering approaches are being explored to reach these goals Tissue engineering approaches Tissue engineering approaches can be divided into three distinct strategies; (1) implantation of decellularized valvular material (implanted as such or re seeded with cells), (2) formation of tissue in vitro by seeding cells onto a biodegradable synthetic or natural scaffold, and (3) implantation of a biodegradable polymer that is remodeled in vivo by endogeneous cells. Implantation of decellularized xenografts or homografts has the advantage that a native like geometry is provided with good mechanical behavior and structure. Varying results have been published on the in vivo functioning of these valves. In sheep, remodeling and growth potential was demonstrated in unseeded decellularized valves without calcification (Erdbrugger et al., 2006; Hopkins et al., 2009). However, in reseeded decellularized valves calcification has been observed (Metzner et al., 2010). Two clinical studies with decellularized valves have been performed in pediatric patients. In the first, decellularized porcine heart valves failed to function due to degeneration and rupture (Simon et al., 2003), while in the other study decellularized human valves re 6

15 General Introduction seeded with endothelial progenitor cells have shown good functionality in pediatric patients up to 3.5 years (Cebotari et al., 2006). Over time, the pulmonary valve diameter increased, while regurgitation decreased, indicating growth of the valve. In adult, pulmonary valve replacements of decellularized xenografts (Konertz et al., 2005) and homografts (Bechtel et al., 2003) have been successful up to two years. The major concerns of decellularized valves are that total cell removal is needed as remnants of cells could induce an immune response, and that preservation of the ECM architecture and composition is required as the disruption of the ECM might reduce the durability of the leaflets. Furthermore, donor valves are required which have a limited availability. The classical tissue engineering approach comprises cell seeded biodegradable scaffolds cultured in vitro in bioreactor systems mimicking native physiological pressures and/or flows. Within this tissue engineering approach many cell sources have been examined (Apte et al., 2011). Neonatal cell sources (Schmidt et al., 2005; Schmidt et al., 2006; Sodian et al., 2006; Schmidt et al., 2007), vascular derived cells (Hoerstrup et al., 2000; Schnell et al., 2001; Mol et al., 2005), endothelial progenitor cells (Sales et al., 2010) and mesenchymal stem cells (Sutherland et al., 2005; Gottlieb et al., 2010) have all shown potential in heart valve tissue engineering. Scaffold materials of synthetic and natural degrading materials have been used. A disadvantage of natural scaffolds, such as fibrin and collagen, is that the polymer itself is weak and long cultures times are required to obtain mechanical integrity (Neidert and Tranquillo, 2006). An alternative to natural scaffolds, are synthetic scaffolds that degrade within a few weeks. Tissue engineered valves have been made based on various combinations of polyglycolic acid (PGA), polylactic acid (PLA), poly 4 hydroxybutyrate (P4HB) and polycaprolactone (PCL). Proof of concept was first demonstrated in 1995 when an autologous tissue engineered leaflet was implanted in sheep (Shinoka et al., 1995). The next step was to develop functional three leaflet tissue engineered valves. Hoerstrup et al., Sodian et al. and Stock et al. were the first to report on in vivo studies with pulmonary autologous seeded tissue engineered heart valves based on synthetic scaffolds. These valves showed functionality at the pulmonary position for up to 24 weeks and were fully remodeled into native like tissues (Hoerstrup et al., 2000; Sodian et al., 2000; Stock et al., 2000). More recent studies have demonstrated further progression in this field. Sutherland et al. have shown in vivo functionality up to eight months (Sutherland et al., 2005), but using a very similar protocol Gottlieb et al. observed increasing regurgitation starting from week six (Gottlieb et al., 2010). Recently, in vivo functionality was demonstrated after eight weeks of implantation with mobile but thickened leaflets (Schmidt et al., 2010). Autologous fibrin based engineered heart valves have demonstrated high quality tissue formation in vitro (Jockenhoevel et al., 2001; Flanagan et al., 2007; Robinson et al., 2008) and recently, an in vivo study presented functionality in sheep (Flanagan et al., 2009). However, the leaflets demonstrated mild shrinkage after three months in vivo, resulting in valvular insufficiency. 7

16 Chapter 1 A relatively new approach of tissue engineering focuses on the direct implantation of scaffold materials into the body without in vitro culture. This approach is based on the hypothesis that the scaffold will be populated with endogenous cells, followed by in vivo tissue formation. The advantage of this in situ approach is that it is cheap, quick and available on demand. However, recruitment of the appropriate cells to remodel the scaffold material is needed. It remains to be determined whether these scaffolds should be pre seeded with freshly isolated cells to attract endogenous cells, or whether this is unnecessary. Recently, freshly isolated autologous bone marrow derived mononuclear cells were seeded onto a biodegradable synthetic scaffold and immediately implanted in primates (Weber et al., 2011). Although this approach was reported to be successful, most of the seeded cells were not present after four weeks in vivo. Roh et al. observed that human bone marrow mononuclear cells, pre seeded into vascular constructs, disappeared within a week, but initiated an inflammation mediated process of vascular remodeling through the attraction of endogenous monocytes. This indicates that pre seeded cells could attract endogenous cells (Roh et al., 2010) Current status of tissue engineered heart valves Although different approaches of heart valve tissue engineering have shown promising results in vivo, none have been adopted in the clinic. The current available heart valve prostheses have their disadvantages, but perform relatively well (15 to 20 years), have predictive behavior and provide a good quality of life. The tissue engineered heart valves need to function properly and provide added value if they are to be accepted by surgeons. Before safe transition to clinical use, some challenges must be overcome in tissue engineering of heart valves. These challenges include thickening of the leaflets in vivo causing loss of flexibility and mimicking the anatomy of the valve and the three layered structure of the leaflet. This thesis will focus on a specific challenge in heart valve tissue engineering, namely valvular regurgitation. During tissue culture, stress is generated within the newly formed tissue due to traction forces exerted by the cells. At release of constraints, this generated stress and the traction forces of the cells cause leaflet shrinkage. When this shortening of the leaflets is severe, closure during diastole is impaired and regurgitation will occur (Flanagan et al., 2009). Valvular regurgitation is observed in almost all in vivo studies, but has always been reported as trivial, minimal, mild and moderate regurgitation (Hoerstrup et al., 2000; Sutherland et al., 2005; Schmidt et al., 2010). However, two recent studies are the first to address leaflet shrinkage as a problem (Flanagan et al., 2009; Gottlieb et al., 2010). Tissue shrinkage is a cell mediated problem, caused by traction forces exerted by the cells. Within the present thesis, this phenomenon of tissue shrinkage is unraveled and both passive and active contributions to it are quantified. Further, the hypothesis that a strong extracellular matrix will 8

17 General Introduction improve the resistance against the traction forces of the cells is investigated. To test this hypothesis, strategies to provide maturation of the collagen network, which provides stiffness and strength to the tissue engineered valves, were investigated. 1.4 Tissue stress, compaction and retraction This thesis focuses on tissue engineered human heart valves fabricated from vascular derived cells, seeded into a PGA scaffold coated with P4HB according to the paradigm illustrated below (Figure 1.4). The cell seeded scaffolds are cultured in a bioreactor system where mechanical stimuli are applied to stimulate tissue growth. After 4 weeks of tissue culture, the scaffold is degraded and a completely autologous living heart valve is grown. Figure 1.4: Schematic overview of in vitro tissue engineering of human heart valves. Human vascular derived cells are expended in the laboratory and seeded onto a heart valve shaped scaffold. Subsequently, the tissues are cultured in a bioreactor system where the valves are subjected to mechanical loading. Finally, the resulting living autologous heart valve would be implanted into the patient. Tissue engineered valves are cultured constrained with the leaflets attached to each other. During culture, stress is generated within the tissue by contractile forces exerted by the cells. This stress has shown to be beneficial for tissue formation and architecture (Mol et al., 2005; Neidert and Tranquillo, 2006; Robinson et al., 2008). However, during culture cell traction mediated remodeling results in compaction of the tissue (Figure 1.5A), represented by leaflet flattening in heart valves (Mol et al., 2005), decreased width in tissue engineered constructs (Shi and Vesely, 2003) and decreased thickness in general (Neidert et al., 2002; Ross and Tranquillo, 2003; Mol et al., 2005). When the constraints are released, tissue retraction occurs due to release of the prestress in the cells and ECM that was generated during culture (passive) and the traction forces exerted by the cells (active) (Balestrini and Billiar, 2009). Tissue retraction is 9

18 Chapter 1 illustrated in figure 1.5B and an example of tissue retraction in an ovine tissue engineered heart valve is presented in figure 1.5C. Tissue compaction and retraction are cell traction mediated processes regulated through cell matrix interactions that transmit information from the extracellular matrix to the cytoplasm and vice versa. Cell matrix interactions regulate cell growth, migration, differentiation, survival, tissue organization and matrix remodeling and are formed by focal adhesion complexes that bind integrin within the cell membrane to the ECM outside the cell (Geiger et al., 2001; Cukierman et al., 2002). Focal adhesions exist in different sizes, related to the amount of stress they exert on the extracellular matrix surrounding them (Goffin et al., 2006). The protein α smooth muscle actin (αsma) is a mechanosensitive protein that generates high contractile activity in stress fibers (Hinz et al., 2001; Hinz et al., 2003). It is present in stress fibers under high tension and closely correlated to focal adhesion size. Therefore, it is often used as a measure for the contractility of a cell (Dugina et al., 2001; Goffin et al., 2006; Hinz, 2006; Wipff and Hinz, 2009). Figure 1.5: Schematic representation of compaction (A) and retraction (B) in tissue engineered strips and heart valves. (A) Compaction causes a decrease in width in tissue engineered strips and leaflet flattening in tissue engineered heart valves. (B) At release of constraints, tissue retraction reduces the length of both tissue engineered strips and leaflets. (C) An ovine tissue engineered heart valve after 4 weeks of culturing with the leaflets attached to each other (I), immediately after separation of the leaflets (release of constraints) (II) and three hours later when kept on ice (III). 10

19 General Introduction In vivo, cells are usually quiescent and become temporarily activated when remodeling is required, associated with increased αsma expression (Rabkin Aikawa et al., 2004). Cell sources used in heart valve tissue engineering generally have an activated remodeling phenotype, which is beneficial for the excessive tissue formation that is needed within a short time frame. However, the accompanying αsma expression (Rensen et al., 2007; Beamish et al., 2010) leads to cell traction mediated problems. Cells need a certain internal stress level to function well and exert traction forces to their surroundings to achieve this. When intracellular tension changes, cells adjust their traction forces to reestablish the preferred internal stress balance (Brown et al., 1998). During culture, cells remodel their environment until they have reached the desired internal stress level, which results in tissue compaction. At release of constraints, the internal stress level of the cells decreases and they will exert traction forces to reach it again, resulting in retraction. Cell traction has been widely investigated in collagen and fibrin gels (Guidry and Grinnell, 1985; Brown et al., 1998; Shi and Vesely, 2003; Balestrini and Billiar, 2009; Chieh et al., 2009). Contractile cell behavior differs between cell sources (Eastwood et al., 1996) and is affected by the boundary stiffness applied to the carrier material (Legant et al., 2009; John et al., 2010). Furthermore, tissue compaction increases when cell density increases and decreases when the density of the provided natural scaffold material increases (Shi and Vesely, 2003; Chieh et al., 2009). To study the evolution of compaction and retraction in human heart valve tissue engineering, the selected constructs in this thesis consist of three components; vascularderived cells, a PGA scaffold coated with P4HB and newly synthesized tissue. The balance between these components changes during culture. The cells within the tissue will always develop a certain internal stress and exert traction forces to their extracellular matrix and scaffold surroundings to achieve this. In the early phase of tissue culture, the scaffold stiffness is high enough to withstand the traction forces of the cells. However, after approximately two weeks of culture, the scaffold rapidly degrades (Klouda et al., 2008), while new tissue is being formed gradually over time. The changes in cell traction, scaffold and tissue quality during culture are illustrated in figure 1.6A. We hypothesize that when the scaffold can no longer withstand the traction forces of the cells, while the newly formed tissue is still weak, an imbalance among the components exists, leading to compaction and retraction. The schematic figure 1.6B illustrates that either slowing down scaffold degradation (Figure 1.6B I) or an increase in tissue stiffness (Figure 1.6B II) can reduce the development of compaction during culture and the resulting retraction. Moreover, temporarily eliminating cell traction forces (Figure 1.6B III) at the time of implantation could reduce tissue retraction. It is worthy of mention that loads applied during the cardiac cycle might be able to counterbalance the traction forces once the valve is functioning in vivo. This, however, can only work if passive retraction is small compared to active retraction. 11

20 Chapter 1 Figure 1.6: (A) Illustration of the hypothesized changes in scaffold and tissue stiffness, and cell traction during four weeks of culture of tissue engineered heart valves. The scaffold degrades rapidly after two weeks, tissue stiffness gradually increases over time, and total traction force of the cells will quickly reach its maximum and will stay at that level. As soon as the scaffold loses its mechanical integrity, while the tissue is still weak, cell traction forces cause compaction and retraction. (B) An overview of hypothesized potential strategies to reduce tissue compaction and retraction. Slowing down scaffold degradation (I) or increasing tissue stiffness (II) make the TE constructs less prone to traction forces, decreasing both compaction and retraction. Reducing cell traction at the time of release of constraints (III) would reduce retraction. 1.5 Collagen maturation Collagen is the main load bearing component of the leaflets, with collagen type I and type III the most abundant types in native heart valves. To provide mechanical integrity and ensure long term in vivo functioning of tissue engineered heart valves, a mature well organized collagen network is needed and thus mimicking native collagen architecture might be necessary. Further, it was hypothesized that a well developed extracellular matrix would be less prone to cell traction forces, reducing tissue retraction at release of the constraints Collagen structure and cross linking Collagen synthesis starts in the endoplasmatic reticulum where three pro αchains are combined into a helical molecule, known as procollagen. This procollagen is transported into the extracellular space where collagen fibrils are formed. Collagen fibrils consist of units of five procollagen molecules staggered together referred to as the microfibrils. Finally, collagen fibrils are organized by the cells into large collagen fibers (Figure 1.7A). Collagen fibrils are stabilized by intermolecular collagen cross links. Mature trivalent collagen cross links are situated between two telopeptides (positioned at the 12

21 General Introduction end of a procollagen molecule) and the triple helix of an adjacent procollagen molecule (Figure 1.7B). Two major forms of mature cross links have been identified, hydroxylysylpyridinoline (HP) and lysylpyridinoline (LP). These cross links are formed by hydroxylation of the telopeptides, where lysine is converted to hydroxylysine, catalyzed by the enzyme lysyl hydroxylase. The hydroxylysine residues serve as a substrate for lysyl oxidase, which transforms some of them into hydroxyallysine. The resulting reactive aldehyde can condense with lysyl or hydroxylysyl into bivalent cross links. Two bivalent cross links can form trivalent HP or LP cross links via a spontaneous chemical reaction. It can take a couple of weeks for this to happen. HP cross links are derived from three hydroxylysyl residues and are predominantly found in highly hydroxylated collagens. LP cross links are derived from two hydroxylysyl and one lysyl residues and are found primarily in calcified tissues. The load bearing capacity of collagen is, apart from collagen content and organization, highly dependent on collagen cross linking, because the cross links stabilize the collagen fibrils (Balguid et al., 2007). Moreover, the presence of mature collagen cross links makes collagen fibers less susceptible to enzymatic degradation (Paul and Bailey, 2003). Figure 1.7: (A) the hierarchical structure of collagen from a fiber down to a triple helix organization (adapted from Biology of the cell, Lewis J., 1994). (B) Schematic of collagen cross links located between two telopeptides and a triple helix In vitro regulation of the collagen network Tissue engineering approaches aim at developing culture methods to improve in vitro collagen synthesis, architecture and maturation. The in vitro development of a wellorganized collagen network can be affected by mechanical, biochemical and environmental stimuli, from which mechanical loading has been investigated most thoroughly. For example, the exposure of developing heart valves to increasing flow and pressure in a bioreactor system improved tissue formation in vitro (Hoerstrup et al., 2000). Strain based mechanical loading of tissue engineered heart valves resulted in 13

22 Chapter 1 superior tissue formation, non linear tissue like mechanical properties (Mol et al., 2005) and native like collagen fiber architecture (Cox et al., 2010). Furthermore, intermittent strain based loading has been shown to accelerate collagen production and cross link formation (Rubbens et al., 2009). Finally, in collagen and fibrin gel based tissue engineered heart valves, commissural alignment of collagen fibers was induced by mechanical constraints (Neidert and Tranquillo, 2006; Robinson et al., 2008). The environmental factor, oxygen (O 2 ), is widely investigated in cancer research and neo vascularization, but has also shown potential to enhance collagen synthesis and cross link formation (Falanga et al., 1993; Agocha et al., 1997; Horino et al., 2002; Brinckmann et al., 2005; Wang et al., 2005). Hypoxia, characterized by low cellular O 2 concentrations, has a strong impact on cell biology. The process behind it is not completely known, but the transcription factor hypoxia inducible factor 1 (HIF 1) is one of the key regulators responsible for the induction of genes in hypoxic conditions. HIF 1 is a complex consisting of a hypoxically inducible subunit HIF 1α and a constitutively expressed subunit HIF 1β. In normoxia, the HIF 1α proteins are rapidly degraded due to hydroxylation of proline into hydroxyproline, resulting in essentially no detectable HIF 1α protein. During hypoxia, HIF 1α remains stable and translocates from the cytoplasm to the nucleus, where it dimerizes with HIF 1β. The HIF 1 complex formed then associates with hypoxia response elements in the regulatory regions of target genes and binds the transcriptional co activators to induce gene expression (Ke and Costa, 2006). Hypoxia has been associated with increased amounts of the collagen cross link enzyme lysyl hydroxylase 2 (Hofbauer et al., 2003; Brinckmann et al., 2005) and procollagen type I and III (Takeda et al., 2000; Chen and Aplin, 2003; Estrada and Chesler, 2009). Furthermore, hypoxia has shown potential in increasing the amount of collagen and collagen cross links in tissue engineered cardiovascular constructs cultured at 7% O 2 (Balguid et al., 2009). 1.6 Rationale and outline Constrained tissue culture, in which stress can be generated by the cells within the tissue, is a requirement for good tissue formation and architecture, but it is also associated with undesired processes such as compaction and retraction. To develop a functional tissue engineered heart valve, it is important to obtain a balance between the beneficial and adverse effects of stress generation within the tissue engineered heart valve. The present thesis focused on the development and adverse effects of stress generation. The aim was: To quantify tissue stress, compaction, and retraction during culture and after release of constraints. To determine the relative contributions of passive and active retraction. 14

23 General Introduction To decrease tissue retraction by improving tissue strength through maturation of the collagen network. First, an in vitro model system was designed to quantify the development of stress and its resulting compaction and retraction in tissue engineered strips (chapter 2). In chapter 3, this model system has been used to unravel the passive and active contributions to tissue retraction. Secondly, improving tissue quality by inducing collagen maturation has been investigated in order to decrease retraction. The potential of the environmental factor oxygen concentration to enhance and improve matrix formation has been investigated at cellular (chapter 4) and tissue level (chapter 5). Finally, the effect of prolonged tissue culture on tissue maturation, stress and retraction was investigated in chapter 6. Finally, the most important findings are discussed in chapter 7, followed by their implications for future research. 15

24 Chapter 1 16

25 Chapter 2 An in vitro model system to quantify stress generation, compaction and retraction in engineered heart valve tissue This chapter is based on: Marijke A.A. van Vlimmeren, Anita Driessen Mol, Cees W.J. Oomens, Frank P.T. Baaijens (2011). An in vitro model system to quantify stress generation, compaction and retraction in engineered heart valve tissues. Tissue Eng Part C Methods. (in press)

26 Chapter Introduction End stage heart valve disease is a frequently occurring disease, most commonly treated with heart valve replacement. Worldwide heart valve replacements are performed annually and this number has been estimated to triple over the upcoming five decades (Yacoub and Takkenberg, 2005; Pibarot and Dumesnil, 2009). Current clinical options for heart valve replacement are bioprosthetic and mechanical valves. Bioprosthetic valves mimic native valve behavior, but are prone to valve degeneration, which makes them less suitable for younger patients (Zilla et al., 2008). Mechanical valves can function for a lifetime, but anticoagulation therapy is needed to prevent thromboembolism (Zilla et al., 2008; Pibarot and Dumesnil, 2009). A tissue engineered heart valve is an autologous and viable prosthesis. Therefore, it is able to repair, remodel and grow throughout a patient s life. Especially for pediatric patients a tissue engineered heart valve would be beneficial, as it would prevent multiple operations to match growth during their childhood. Autologous tissue engineered heart valves are fabricated by seeding extracellular matrix producing cells on a degradable scaffold material and culture this in vitro. During the culture period the scaffold material degrades and the extracellular matrix produced by the cells takes over. In vivo results of autologous heart valves have shown promising results up to 8 months (Sodian et al., 2000; Sutherland et al., 2005; Gottlieb et al., 2010). Arterial or venous derived cells are commonly used as cell sources for this approach (Schnell et al., 2001). These cells are associated with high extracellular matrix production, but their contractile properties may be a source of problems. In our group, tissue engineered heart valves are grown with the leaflets attached to each other to assure constraint tissue culture, which has shown to be beneficial for tissue formation (Mol et al., 2005). In this constrained situation, the cells are able to generate stress within the tissue, due to their contractile properties. Just before implantation, the leaflets are separated to enable opening and closing of the valve in vivo. The stress that has been build up within the tissue during culture causes retraction of the leaflets once constraints are released (Figure 2.1). If this retraction is severe, the heart valve will not be able to fully close during diastole and its functionality is lost. Tissue retraction is a serious issue within heart valve tissue engineering as dysfunctional closing of the valve leaflets leads to regurgitation. Valvular regurgitation, related to decreasing leaflet length, has been observed in both natural and synthetic scaffold based autologous heart valves (Flanagan et al., 2009; Gottlieb et al., 2010). Tissue retraction results from an imbalance among the three major components of engineered heart valve tissues: the cells, the scaffold and the newly formed tissue. The cells within the tissue will always aim to develop a certain internal stress (Brown et al., 1998) and exert traction forces to the extracellular matrix (ECM) and scaffold surroundings to achieve this. In the early phase, the scaffold is stiff enough to withstand the contractile forces of the cells. However, after approximately 2 weeks of culture, the 18

27 Stress generation, compaction and retraction scaffold rapidly degrades, while the newly formed tissue is not capable yet to withstand the contractile forces. The changes in cell traction, scaffold and tissue during culture are illustrated in figure 2.2. As a result of these changes, two processes of tissue shrinkage can be distinguished. During culture, tissue compaction occurs by cell traction mediated remodeling of the construct. Upon release of the constraints, retraction occurs immediately due to release of the stress generated within the matrix by the cells during culture and slowly continues due to retraction of the matrix by the cells (Balestrini and Billiar, 2009). Figure 2.1: An ovine tissue engineered heart valve after 4 weeks of culturing with the leaflets attached to each other (A), immediately after separation of the leaflets (B) and 3 hours later when kept on ice (C). To solve the problem of tissue shrinkage, it is important to know how the imbalance between scaffold, tissue and cell traction results in stress generation, compaction and tissue retraction. Therefore, the aim of the present study is to quantify and correlate tissue compaction, retraction and stress generation in engineered heart valve tissue during a culture period of 4 weeks. We developed an in vitro model system in which stress generation, compaction and retraction can be measured in a single engineered tissue. Proof of principle of this model system is studied by culturing engineered heart valve tissues based on human vascular derived cells seeded onto a rapid degrading scaffold within this model system for 4 weeks. Stress generation, compaction and retraction were quantified during culture and after release of constraints at week 4. Using this model system, we are able to gain insight in the development of stress, needed to unravel the balance between tissue, scaffold and cell traction. Therefore this model system is of great value for investigating the applicability of possible solutions to prevent or compensate for tissue retraction towards a functional tissue engineered heart valve. 19

28 Chapter 2 Figure 2.2: Illustration of the changes in scaffold and tissue stiffness, and cell traction during 4 weeks culture of tissue engineered constructs. The scaffold degrades rapidly after two weeks, tissue stiffness gradually increases over time, while total traction force of the cells will quickly reach its maximum and will stay at that level. 2.2 Material & Methods Experimental set up An in vitro model system is developed in which generated stress, compaction and retraction of one tissue engineered construct can be quantified during culture and after release of constraints (Figure 2.3A). The model system consists of a stainless steel frame in which two ultra high molecular weight polyethylene (UHMWPE) sliding blocks are positioned opposite from each other. In between the two sliding blocks, the tissue engineered constructs (TE constructs) are cultured. One sliding block can be either fixed (during culture) or move freely (to measure retraction after culture, Figure 2.3B). The other is connected to the metal frame via two leaf springs and the displacement of this sliding block is related to the generated force within the tissue (Figure 2.3C). The sliding blocks can be fixed with clamps (Figure 2.3A 1,2). To measure force generation during culture, the sliding block of the leaf springs is not fixed. Because the force that is generated by the cells will deform the leaf springs this configuration is called semiconstrained. To measure generated force after release of constraint, both sides are fixed and this configuration is called constrained. 20

29 Stress generation, compaction and retraction Figure 2.3: Photographs (ABC 1) and a schematic overview (ABC 2) of model system in which retraction ( R L, B1 2) and generated force ( F, C1 2) can be measured through the displacement ( v ) of two sliding blocks positioned opposite from each other. (1) Clamp to fixate retraction side. (2) Clamp to fixate force side. (3) Dots to measure retraction. (4) Dots to measure generated force. L0 and L retr represent the original and retracted length of the TE constructs. Both generated force and retraction are measured through the displacement of the sliding blocks by comparing the displacement of a black dot on the sliding block to a reference dot on the metal frame. To quantify the displacement during culture (generated force), photographs were made twice a week and analyzed with ImageJ (WCIF ImageJ, National Institutes of Health, USA). Displacements at week 4 (retraction and generated force after release of constraints) were assessed from pictures taken underneath a stereomicroscope (Zeiss Observer, Zeiss, Göttingen, Germany) and quantified with Matlab (The MathWorks, Eindhoven, The Netherlands) Quantification of the generated stress Theoretically, the force generated by the tissue engineered construct (mn) can be calculated from the displacement v of the sliding block (mm) and the properties of the leaf springs from the equation for bending of a straight beam (Young et al., 2002): vei F 24 (2.1) L 3 where E is the young s modulus of the beam material (mn/mm 2 ), is the moment of inertia determined from the width b (mm) and height h (mm) of the leaf spring 3 ( I 1 bh,mm 4 ), and L is the length of the leaf springs (mm). For this model system, E 12 was 2x10 2 nn/mm 2, I was 1.04x10 5 mm 4 and L was 14 mm. Since the leaf springs were made and positioned by hand, each model system was calibrated with an individual fit: F slope v (2.2) 21

30 Chapter 2 This was done by measuring the displacement of the sliding block during three cycles of subjection to known applied forces ranging from 0 to 20 mn. Loading was within the linear region of the leaf springs verified by the linearity of the force displacement curves. To translate the generated force to stress (kpa), the cross sectional area A (mm 2 ) of the TE constructs was measured. Width was assessed during culture from photographs and underneath a stereomicroscope at week 4. The initial thickness at day 0 was 1 mm, representing the thickness of the scaffold. Thickness at week 4 was assessed from histological sections (see under tissue formation), assuming a linear shrinkage factor of during processing of the tissue for histology (Schned et al., 1996). Thickness during culture could not be measured and, therefore, the thickness at earlier time point was interpolated assuming a change over time with a course similar to that of the width. The calculated cross sectional area was used to determine the generated stress via: F (2.3) A Tissue culture Vascular derived cells were harvested from the human vena saphena magna obtained according to the Dutch guidelines for secondary used materials. These cells have previously been characterized as myofibroblasts (Mol et al., 2006). They show expression of vimentin, but not desmin and a subpopulation of the cells express α smooth muscle actin. Cells were expanded using standard cell culture methods in a humidified atmosphere containing 5% CO 2 at 37 C. Culture medium consisted of advanced Dulbecco s Modified Eagle Medium (DMEM; Invitrogen, Carlsbad, USA), supplemented with 10% Fetal Bovine Serum (FBS; Greiner Bio one, Frickenhausen, The Netherlands), 1% GlutaMax (Invitrogen), and 1 % penicillin/streptomycin (Lonza, Basel, Switzerland). The cells were seeded onto rectangular shaped scaffolds (18x5x1 mm) of rapid degrading nonwoven polyglycolic acid (PGA) (Concordia Manufacturing Inc, Coventry, RI, USA) coated with poly 4 hydroxybutyrate (P4HB) (Tepha, Lexington, Massachusetts; received as part of the collaboration with University Hospital Zurich). The scaffolds were attached to the sliding blocks with polyurethane tetrahydrofuran glue (15% wt/vol). Sterilization was achieved by 70% ethanol incubation for 30 minutes. Cells were seeded at passage 7 with a seeding density of 15 million cells per cm 3 using fibrin as a cell carrier. The TE constructs were cultured in rectangular well plates for 4 weeks in a humidified atmosphere containing 5% CO 2 at 37 C. During tissue culture, the medium was supplemented with L ascorbic acid 2 phosphate to promote extracellular matrix production (0.25 mg/ml; Sigma, St. Louis, MO, USA), and replaced twice a week. 22

31 Stress generation, compaction and retraction Experimental design TE constructs were cultured for 4 weeks in constrained (n=5) and semiconstrained (n=5) configuration. During culture, the compacted width W comp (µm) of both constrained and semi constrained TE constructs (n=10) was assessed twice a week. Compaction C W (%) was defined as the percentage of shrinkage compared to the original widthw 0 : Wcomp 1 100% C (2.4) W Stress generation during culture was measured in the semi constrained TE constructs twice a week. After 4 weeks of culturing, generated stress and retraction were measured in the constrained samples after release of the constraints. First, generated stress was measured at 3 minute intervals over a total period of 30 minutes, after which retraction in length R (%) was measured during 24 hours as the percentage L of shrinkage compared to the original length L (µm): 0 W 0 Lretract 1 100% R (2.5) L in which L is the retracted length (µm) of the TE construct. Retraction was measured retract every 6 minutes in the first half hour, followed by measurements after 1, 2, 4, 6, 16 and 24 hours. An overview of the experimental design is given in figure 2.4. L 0 Figure 2.4: Schematic overview of the experimental design. Compaction during culture was measured in both constrained and semi constrained samples. Retraction and generated stress were measured at week 4 after release of constraints in the constrained samples. Generated stress during culture was measured in the semi constrained samples. 23

32 Chapter Tissue formation Tissue quality was evaluated by histological staining. The TE constructs were fixed in 3.7% formaldehyde (Merck) and embedded in paraffin. Tissue sections of 10 µm were cut and studied by hematoxylin and eosin staining for general tissue morphology and Masson Trichrome staining for deposition of collagen. The stained sections were evaluated using light microscopy (Axio Observer, Zeiss, Göttingen, Germany). Additional sections were stained with Picrosirius red to visualize mature collagen fibers and evaluated using polarized light microscopy Statistical analyses All data are presented as mean and their standard error of mean. Linear regression analysis was performed to fit the calibration curves using GraphPad Prism software (GraphPad Software, Inc, USA). 2.3 Results Calibration of the model systems Given the constants of the materials and the model system (see quantification of generated stress), the theoretical correlation between force ( F in mn) and displacement ( v in µm) becomes F v. Figure 2.5 shows a representative fit and its 95% confidence interval for one of the 20 model systems that were calibrated. The average slope of all model systems was 0.021±0.003, indicating that bending of the leaf springs is slightly stiffer than theoretically determined. The averaged 95% confidence interval of all model systems comprised 8.7% of the generated force. Figure 2.5: Representative force displacement calibration curve. The dotted lines indicate the 95% confidence intervals. 24

33 Stress generation, compaction and retraction Tissue culture in the model system TE constructs were successfully cultured in the model system for 4 weeks in both constrained and semi constrained configuration (Figure 2.6A C). Newly formed tissue and compaction in width was clearly visible at week 4 (Figure 2.6B,C) compared to day 0 (Figure 2.6A). In figure 2.6C it can be seen that the generated stress has displaced the sliding block attached to the leaf springs. The hematoxylin and eosin staining showed homogeneous tissue formation (Figure 2.6D). The Masson Trichrome and Picrosirius red staining indicate a dense collagen network with mature collagen fibers (Figure 2.6D F) after 4 weeks. Figure 2.6: Tissue engineered constructs just after seeding (A) and after 4 weeks of culture in constrained (B) and semi constrained (C) configuration. Masson Trichrome (D), Hematoxylin and Eosin (E) and Picrosirius Red (F) staining showed homogeneous tissue formation with a dense and mature collagen network. The white bars represent 200 µm Tissue compaction and generated stress during culture The TE constructs started to compact after 2 weeks and continued to compact to approximately half of their original width (C W = 51.2±4.8%) at week 4 (Figure 2.7A). At day 0, tissue thickness was 1000 µm and at week 4 a thickness of 1067 µm was assessed from histology. It was assumed that tissue thickness increase over time changed to a pattern similar to that of width decrease, starting after 2 weeks (Figure 2.7B). Using this assumption, the total cross sectional area of the TE constructs was determined, which compacted from 5.44±0.08 mm 2 at start to 2.57±0.17 mm 2 after 4 weeks (Figure 2.7C). Force and stress generation showed the same time lapse as compaction, starting after 2 25

34 Chapter 2 weeks and gradually increasing up to respectively 20.3±2.9 mn and 8.0±1.3 kpa at week 4 (Figure 2.8). Figure 2.7: Tissue compaction (A) started after two weeks and continued up to week 4. Tissue width (B, filled triangles) was measured during culture and it was assumed that tissue thickness (B, open triangles) followed the same course to the measured value at week 4. These values for thickness and width resulted in the decreasing cross sectional area over time illustrated in (C). Figure 2.8: Force (A) and stress (B) generation started at week 2 and continued to increase up to respectively 20.3±2.9 mn and 8.0±1.3 kpa at week 4. 26

35 Stress generation, compaction and retraction Tissue retraction and generated stress after release of constraints Within 5 seconds after release of constraints, the measured generated force and stress were respectively 8.7±0.5 mn and 3.3±0.3 kpa. After 30 minutes, generated force had gradually increased to 15.1±0.7 mn, which correlated to a generated stress of 5.6±0.6 kpa (Figure 2.9A,B). Next, retraction was measured which occurred very fast in the first 20 minutes after release of constraints. This resulted in a retraction of 16.4±1.3%, which was almost half of the total retraction after 24 hours being 35.8±3.3% (Figure 2.9C). Figure 2.9: Tissue retraction (A) occurred rapidly in the first 20 mintes after which is gradually continued up to 35.8±3.3% after 24 hours. Generated force (B) and stress (C) reached maximum values of 15.1±0.7mN and 5.6±0.6kPa after 30 minutes. 2.4 Discussion This paper describes a new in vitro model system for investigating the development of stress, generated by traction forces exerted by the cells, in TE constructs. This cell traction on the one hand is beneficial for tissue maturation and alignment in engineered tissues (Mol et al., 2005; Neidert and Tranquillo, 2006; Robinson et al., 2008), while on the other hand is causing tissue shrinkage at release of constraints (Figure 1). Tissue shrinkage results from an imbalance between scaffold and tissue stiffness, and the traction forces of cells (Figure 2). When traction forces exerted by the cells are larger than the forces the scaffold or newly formed tissue can resist, 27

36 Chapter 2 tissue shape will change. During tissue culture, scaffold and tissue properties change, influencing the effect of cell traction forces on tissue shape. In the present study, a model system is developed in which stress generation, compaction and retraction can be measured in one single engineered heart valve tissue. Histology confirmed that the model system is suitable for tissue culture showing homogeneous tissue and abundant collagen formation after 4 weeks of culture Tissue compaction and generated stress during culture The TE constructs started to compact after 2 weeks of culture and continued to compact to approximately half of their original width at week 4. Generated force was quantified with an averaged error of 8.7% and divided by the cross sectional area to obtain stress. The stress, generated by the cells, followed the same time course as compaction reaching a value of 8.0 kpa at week 4. However, at week 4, compaction seemed to level off, while no signs of leveling off were observed in the stress. Tissue compaction has been investigated extensively in all sorts of cell populated constructs, predominantly made of collagen type I gels (Guidry and Grinnell, 1985; Shi and Vesely, 2003; Stegemann and Nerem, 2003; Chieh et al., 2009). Constrained culture of smooth muscle cells in collagen gels resulted in a compaction of 90% in 21 days (Shi and Vesely, 2003), while faster compaction of 98% in 6 days was observed in unconstrained disk gels (Stegemann and Nerem, 2003). Bone marrow stromal cells in constrained collagen lattices compacted 90% in 3 days time (Chieh et al., 2009) and human dermal fibroblasts seeded on top of collagen gels reached a compaction in thickness of 85%, while being constrained to the surface (Guidry and Grinnell, 1985). Compaction of constrained tissues made from cell sheets of human dermal fibroblast was less and slower, reaching a compaction of 70% after 21 days of constrained culture (Grenier et al., 2005). Contrary to our findings, compaction in most of these studies occurs within hours to days and the amount of compaction is higher ranging from 50% up to 90%. These differences are probably caused by the type of scaffold used, its degradation profile and the formation of new tissue. During the first 2 weeks, the traction forces of the cells are most likely counterbalanced by the stiffness of the scaffold. After 2 weeks, the scaffold rapidly looses its mechanical integrity (Klouda et al., 2008) and the traction forces of the cells can compact the construct, which explains why first compaction was observed after 2 weeks. The fact that compaction does not reach extreme levels up to 90% as observed in collagen lattices implies that the developed ECM is less prone to the traction forces of the cells than a collagen gel is. It has been shown that increasing collagen concentration leads to decreased compaction (Shi and Vesely, 2003; Chieh et al., 2009). We believe that a fully developed matrix limits compaction as its collagen network is denser and stronger (due to cross links) than the collagen network of a collagen gel. 28

37 Stress generation, compaction and retraction Studies of human dermal fibroblasts in collagen lattices have shown that cells exert traction forces within hours to reach tensional homeostasis, and that these forces level off after 6 36 hours (Eastwood et al., 1996; Brown et al., 1998; John et al., 2010). In the present study, 2 weeks were needed to generate stress and no plateau was reached at week 4. This phenomenon can also be explained by the degradation of the scaffold and the development of tissue over time. However, it is expected that if culture time is increased, the tissue further develops and equilibrium between traction forces and tissue stiffness would eventually be reached in the studied TE constructs as well Tissue retraction and generated stress after release of constraints Retraction at release of constraints occurred very fast in the first 20 minutes, after which it gradually continued up to 36% after 24 hours. The measured force also consisted of a fast and slow part similar to retraction. Immediately after release of constraints, a force of 8.7mN was measured. After that, the measured force kept increasing and reached a value of 15.1mN after 30 minutes. The fast processes of retraction at release of constraints are caused by the release of the developed stress in the ECM during culture and the release of stress in the cells. Next to this passive retraction, cells will also actively exert traction forces on their environment when their internal stress is released, seeking to reach their basal internal tension again. This process results in further retraction and measured force, but in a slower fashion. Although no new constraints are enforced to the tissue, retraction is hypothesized to stop eventually due to volume restrictions. Compared to fibroblast populated fibrin gels, retraction in the TE constructs was slow. Fibroblast populated fibrin gels retracted 60% after release from the surface within 10 minutes (Balestrini and Billiar, 2009), even though cell density was 30 times lower than in the present experiment. This implies again that a fully developed ECM is better in withstanding traction forces of cells than a sole gel is Limits and future applications of the developed model system Translation of force to stress required the tissue cross sectional area. Width was measured in all samples but a histology section was used to estimate tissue thickness at week 4. A tissue shrinkage factor of was assumed to correct for tissue dehydration during preparation and the thickness at earlier time points was determined by interpolation. A more accurate method should be developed for future experiments as both the tissue shrinkage factor and interpolation over time could be sources of errors. However, in this experiment changes in thickness appeared to be less than changes in width. Therefore, thickness changes had a minimal influence on the translation from force to stress. 29

38 Chapter 2 The generated force and stress were slightly higher in the semi constrained samples than in the constrained ones. In a fully constrained construct, cells will remodel their environment until an internal equilibrium in stress level is reached. In the semiconstrained samples, a certain degree of compaction (due to bending of the leaf springs) as well as generation of stress will occur as the cells exert traction forces to the tissue. It is hypothesized that this possibility of matrix remodeling in length leads to enhanced force and stress generation. Although the amount of compaction was only small in the semi constrained samples, it does affect the position of the sliding blocks at which the cells reach their desired intracellular tension. The difference between constrained and semi constrained can be reduced by increasing the stiffness of the leaf springs. However, increasing the stiffness of the leaf springs will reduce the displacement of the sliding block and therewith the accuracy of the force measurement. The TE constructs are fabricated with seeding techniques, cells and scaffold material similar to those of TE heart valves and therefore are a good representative of TE valves. The semi constrained configuration resembles the situation within a leaflet more closely than the constrained configuration, as during valve culture leaflet flattening occurs due to cell traction. Retraction of the leaflets at release of constraints does differ from the TE constructs as it is counteracted by the constraints in the perpendicular direction. These differences should be investigated, i.e. with computational models, when translating these results to TE heart valves. The development of the currently described in vitro model system enables us to study and resolve the problem of retraction by fine tuning the balance between scaffold, tissue and cell traction. As shown by the time course of stress development in this study, the scaffold is able to withstand compaction during the first two weeks. After that, tissue has to take over, but is not stiff enough yet. This suggests that either decreasing scaffold degradation (Figure 2.10 A) or increasing tissue stiffness (Figure 2.10 B) would change the course of compaction during culture and the resulting retraction. Other approaches could focus on reducing cell traction forces (Figure 2.10 C), by for example disruption of the actin fibers of the cells, disruption of cell matrix interaction or the use of a less contractile cell source (Figure 2.10 D). In addition to these problem solving approaches, the developed model system is also a tool for further investigation of the role of cellular elements in compaction and retraction. Distinguishing between active and passive processes is required for understanding and controlling the process of retraction and is subject of future studies. Finally, in vivo, heart valves are subjected to continuous loading, which will counteract with the traction forces of the cells. This in vivo load could perhaps be used in finding the balance between cell traction and shape conservation. Now the force exerted by the cells is known, one can estimate the load needed to withstand retraction. 30

39 Stress generation, compaction and retraction Figure 2.10: An overview of strategies to reduce tissue compaction and retraction. Decreasing scaffold degradation (A) or increasing tissue stiffness (B) would make the TE constructs less prone to traction forces, decreasing both compaction and retraction. Reducing cell traction at time of implantation (C) would reduce retraction. Using a less contractile cell source (D) could reduce both compaction and retraction Conclusion In conclusion, with the developed model system we were able to quantify generated stress, compaction and retraction over time and after release of constraints. In engineered heart valve tissue, both compaction and stress generation started after 2 weeks of culture and continued up to week 4. TE constructs compacted up to half of their original width and developed an internal stress of 6 8 kpa at week 4, which resulted in a retraction of 36%. The model system has provided a useful tool to unravel and optimize the balance between traction forces of the cells, tissue and scaffold properties in engineered tissues to develop functional tissue engineered heart valve leaflets. 2.5 Acknowledgements The authors gratefully acknowledge the support of the Smart Mix Program of the Netherlands Ministry of Economic Affairs and the Netherlands Ministry of Education, Culture and Science. The authors would like to thank Rob van den Berg for his contributions to the development of the model system 31

40 Chapter 2 32

41 Chapter 3 Passive and active contributions to generated force and retraction in heart valve tissue engineering This chapter is based on: Marijke A.A. van Vlimmeren, Anita Driessen Mol, Cees W.J. Oomens, Frank P.T. Baaijens (2011). Passive and active contributions to generated force and retraction in heart valve tissue engineering. Biomech Model Mechan. (submitted)

42 Chapter Introduction Heart valve tissue engineering is a widely investigated technology in the field of regenerative medicine as an alternative for current available heart valve replacements. Tissue engineered heart valves provide viable implants, which may have the capacity to grow, remodel and repair throughout a patient s life. For pediatric patients this would prevent dangerous re operations throughout their childhood to accommodate growth. Over the last decade, bioreactor systems inducing mechanical conditioning and flow profiles have improved in vitro tissue formation (Mol et al., 2005; Flanagan et al., 2007; Kortsmit et al., 2009; Ruel and Lachance, 2009; Syedain and Tranquillo, 2009). Moreover, in vivo animal studies have demonstrated the feasibility of heart valve tissue engineering, showing remodeling into native like structures (Hoerstrup et al., 2000; Sodian et al., 2000; Stock et al., 2000; Sutherland et al., 2005; Flanagan et al., 2009; Gottlieb et al., 2010; Schmidt et al., 2010). However, there are still some challenges that need to be overcome. One of those challenges is cell mediated leaflet retraction, causing regurgitation in vivo (Flanagan et al., 2009; Gottlieb et al., 2010). In general, this regurgitation has been neglected and reported to be only mild. However, it is observed in every in vivo study and should be considered a serious issue that needs to be addressed. Leaflet retraction is a cell mediated problem, caused by traction forces that the cells exert on their surroundings. Cells will always aim to develop an internal stress homeostasis and adjust their traction forces to retain this in response to environmental changes. Increased traction forces are observed upon a decrease of mechanical loading, while a decrease is observed upon increased mechanical loading (Brown et al., 1998; Mizutani et al., 2004). Within the field of heart valve tissue engineering, the effects of these cell traction forces are controversial. During constrained tissue culture, the traction forces of the cells allow them to remodel their surroundings, which results in compaction of the tissue and the generation of pre stress within the extracellular matrix (ECM) (Balestrini and Billiar, 2009; Van Vlimmeren et al., 2011a). This stress development within the cells and matrix is beneficial for tissue formation (Mol et al., 2006) and provides an oriented collagen network (Neidert and Tranquillo, 2006; Robinson et al., 2008). However, upon release of constraints the developed pre stress in the ECM and cells causes immediate retraction of the tissue (passive process), after which the cells start to remodel their surroundings in an attempt to reestablish their desired internal stress level (active process), leading to additional retraction (Balestrini and Billiar, 2009). To prevent leaflet retraction in heart valve tissue engineering, these passive and active processes need to be unraveled and quantified. Cell traction forces are transmitted from the intracellular network of actin fibers to the surrounding ECM via focal adhesions that consist of complex anchoring proteins, such as aggregated integrins, vinculin, paxin, tensin and focal adhesion kinase (Dugina et al., 2001; Hinz and Gabbiani, 2003; Hinz, 2006). Intracellular, traction forces are 34

43 Passive and active contributions to generated force and retraction generated by actin myosin contraction, which is promoted by phosphorylation of myosin light chain (MLC). In myofibroblasts, this actin myosin contraction is mediated by the Rho A and Rho associated kinase (ROCK) pathway, which blocks MLC phosphatase and directly phosphorylates MLC (Yee et al., 2001; Dallon and Ehrlich, 2010; Follonier Castella et al., 2010; Nobe et al., 2010). Traction forces can be blocked by the disruption of actin fibers or inhibition of the ROCK pathway, and increased by enhancement of the serum concentration in the culture medium (Wakatsuki et al., 2000; Grouf et al., 2007; Nobe et al., 2010). Blocking traction forces, the active cellular component of retraction can be eliminated. Active and passive phenomena have been investigated thoroughly in short term studies with fibroblast populated collagen and fibrin gel systems (Tomasek et al., 1992; Grouf et al., 2007; Balestrini and Billiar, 2009; Legant et al., 2009; Marquez et al., 2009; John et al., 2010; Youssef et al., 2011). In tissue engineered heart valves, the ECM is almost fully developed during 4 weeks of tissue culture and is more mature than the generally investigated collagen and fibrin gels. Tissue retraction and remodeling is less easy in such a developed tissue, as the surroundings of the cells are stiffer. Therefore, this study quantified the passive and active contributions to generated force and retraction in engineered heart valve tissues. This was performed in a model system of tissue engineered constructs, fabricated from vascular derived cells seeded onto a biodegradable scaffold. First, the active cellular traction forces were eliminated by the disruption of the actin fibers with Cytochalasin D (CytoD) and the inhibition of the ROCK pathway by Y As force and retraction remained substantial, total cellular contribution (both passive and active) was eliminated next, by lysis of the cells and decellularization of the tissue. Furthermore, maximum retraction was investigated by increasing the concentration of serum. 3.2 Material & Methods Tissue culture Human vascular derived cells were harvested from the vena saphena magna obtained according to the Dutch guidelines for secondary used materials. Cells were expanded using standard cell culture methods in a humidified atmosphere containing 5% CO 2 at 37 C. Culture medium consisted of advanced Dulbecco s Modified Eagle Medium (DMEM; Invitrogen, Carlsbad, USA), supplemented with 10% Fetal Bovine Serum (FBS; Greiner Bio one, Frickenhausen, The Netherlands), 1% GlutaMax (Invitrogen), and 1% penicillin/streptomycin (Lonza, Basel, Switzerland). Rectangular tissue engineered strips (referred to as TE constructs) were cultured in a previously described model system (Van Vlimmeren et al., 2011a). Briefly, the TE 35

44 Chapter 3 constructs were cultured in between two sliding blocks positioned in a metal frame. One of the sliding blocks was fixed to the metal frame via two leaf springs and the displacement of this sliding block correlated to the generated force. Because the force generated by the cells during culture deformed the leaf springs, it was defined as semiconstrained culture. The opposite sliding block was fixated during culture and used to measure retraction after release of the constraints at the end of the culture period. To fabricate the TE constructs, the cells were seeded onto rectangular shaped scaffolds (18x5x1 mm) of rapid degrading nonwoven polyglycolic acid (PGA) (Concordia Manufacturing Inc, Coventry, RI, USA) coated with poly 4 hydroxybutyrate (P4HB, received as part of the collaboration with the University Hospital Zurich) as described previously (Mol et al., 2005). In summary, the scaffolds were attached to the two sliding blocks with polyurethane tetrahydrofuran glue (15% wt/vol). Sterilization was achieved by 70% ethanol incubation for 30 minutes. Cells were seeded at passage 7 with a seeding density of 15 million cells per cm 3 using fibrin as a cell carrier (Mol et al., 2004). The TE constructs were cultured for 4 weeks. During tissue culture, the medium was supplemented with L ascorbic acid 2 phosphate (0.25 mg/ml; Sigma, St. Louis, MO, USA) and replaced twice a week Force generation and retraction Both force generation and retraction were quantified in Matlab (The MathWorks, Eindhoven, The Netherlands) based on the displacement of the sliding blocks, using photographs made underneath a stereomicroscope (Zeiss Observer, Zeiss, Göttingen, Germany) (Van Vlimmeren et al., 2011a). Every model system was calibrated with its own force displacement curve. Retraction was defined as the percentage of shrinkage compared to the length of the TE construct at time of release of the constraints Experimental design After 4 weeks of culture, the generated force was quantified based on the displacement of the leaf springs compared to day 0. Changes in force and retraction were compared to control TE constructs that did not undergo any treatment. In the first experiment, the active cellular contribution to the generated force and retraction was investigated (n=4 8 per group). The TE constructs were incubated 2 hours with 50% FBS to maximize cellular activity, 10µM cytochalasin D (Sigma) to disrupt the actin network of the cells and 10µM Y (Sigma) to inhibit the ROCK pathway. After incubation with these biochemicals, the change in force was quantified. Next, constraints were released and tissue retraction was measured during 2 hours with time intervals of 15 minutes. Subsequently, the TE constructs were washed twice with PBS 36

45 Passive and active contributions to generated force and retraction and received fresh tissue culture medium, except for the FBS samples which obtained fresh 50% FBS medium. Then, further retraction was measured to examine the recovery from CytoD and the ROCK inhibitor, and maximum retraction in the FBS group. During the first 8 hours, retraction was measured every 30 minutes, followed by measurements at 24 and 48 hours. As retraction was still substantial in this first experiment, we performed a second experiment in which cellular contribution was completely eliminated by lysis of the cells and complete decellularization of the tissue (n=5 per group). Force was again measured before and after treatment, followed by retraction measurements. Retraction was measured every 15 minutes during the first 2 hours, followed by every 30 minutes up to 10 hours and one final measurement after 24 hours Lysis and decellularization Lysis and decellularization of the TE constructs was performed after 4 weeks of tissue culture. The TE constructs were washed three times with PBS, after which the cells were lysed by incubation with a detergent solution (0.25% triton X 100 (Merck, Schiphol rijk, The Netherlands) % sodium deoxycholate (Sigma) % EDTA (Sigma)) over night at 37 C. The lysed samples were then washed in PBS and placed in medium to measure force and retraction. For the decellularized samples, cellular removal was established by nuclease treatments with Benzonase Nuclease (Novagen, EMD Chemicals Inc., San Diego, USA). The TE constructs were incubated with subsequently 100 U/ml (5 8 hours), 80 U/ml (over night) and 20 U/ml (5 8 hours) of Benzonase in 50 mm TRIS HCL buffer (Sigma). Finally, they were washed with PBS and placed in medium to measure force and retraction Immunocytochemistry and immunohistochemistry The effect of 50% FBS, CytoD and ROCK inhibitor on the actin network of the cells was evaluated by a whole mount F actin staining (n=2 per group). In brief, after 2 hours of incubation with the biochemicals, the TE constructs were washed three times with PBS for 5 minutes, followed by fixation with 3.7% formaldehyde (Merck) for 40 minutes. Next, samples were permeabilized with 0.5% Triton X 100 (Merck) for 30 minutes. Subsequently, the TE constructs were incubated with Phalloidin TRITC (1:500, Sigma) for 60 minutes, after which they were washed three times with PBS for 5 minutes. Finally, the TE constructs were incubated with 4,6 diamidino 2 phenylindole (DAPI) for 15 minutes, washed three times with PBS for 5 minutes and evaluated using a multiphoton microscope (Zeiss LSM 510 META NLO, Darmstadt, Germany) in Two Photon LSM mode. 37

46 Chapter 3 The efficacy of cellular lysis and tissue decellularization were evaluated qualitatively by histological and immunofluorescent stainings (n=1 per group). The TE constructs were fixated with 3.7% formaldehyde (Merck), embedded in paraffin and cut in tissue sections of 10 µm. Samples were studied by hematoxylin and eosin (HE) staining for general tissue morphology and cellular content. Further, a DAPI staining was performed to check the presence or absence of cell nuclei. The HE staining was visualized using light microscopy (Axio Observer, Zeiss, Göttingen, Germany) and the DAPI staining was evaluated using fluorescence microscopy (Axiovert 200M, Zeiss, Göttingen, Germany) Biochemical composition of tissues The amount of DNA was assessed from all samples (n=56). In control, lysed and decellularized samples, the amount of sulfated glycosaminoglycans (GAGs) and collagen were determined as well to investigate the effect of lysis and decellularisation on the ECM composition (n=4 per group). Lyophilized tissue samples were digested in papain solution (100 mm phosphate buffer, 5 mm L cysteine, 5mM ethylenediaminetetraacetic acid (EDTA), and µg papain per ml) at 60 C for 16 hours. DNA content was determined with the Hoechst dye method (Cesarone et al., 1979) and a standard curve of calf thymus DNA (Sigma). GAG content was determined with a modification of the assay described by Farndale et al. (Farndale et al., 1986) and a standard curve from chondroitin sulfate from shark cartilage (Sigma). In short, 40 μl of diluted sample, without addition of chondroitin AC lyase, chondroitin ABC lyase and keratanase, was pipetted into a 96 well plate in duplicate. Subsequently, 150 μl dimethylmethylene blue was added and absorbance was measured at 540 nm. The hydroxyproline content was determined with an assay according to Huszar et al. (Huszar et al., 1980) and a reference of trans 4 hydroxyproline (Sigma). A 1 to 8.9 ratio of hydroxyproline to collagen was assumed Statistical analyses All data are presented as mean and their standard error of mean. One way ANOVA, followed by a Tukey s multiple comparison post hoc test, was carried out to evaluate statistical differences. Statistical analyses were done using GraphPad Prism software (GraphPad Software, Inc, USA) and considered significant for P values <

47 Passive and active contributions to generated force and retraction 3.3 Results The effect of 50% FBS, Cytochalasin D and ROCK inhibitor on force and retraction Figure 3.1 gives an overview of representative immunofluorescent stainings of the actin fibers within the cells at 14 and 28 µm tissue depths. Both control and FBS samples contained aligned and organized actin fibers. Incubation with CytoD disrupted the actin network. In the superficial surface layer (14 µm) cells became completely rounded, while stretched cells with a fragmented actin network were observed deeper in the tissue (28 µm). Incubation with ROCK inhibitor was less destructive for the actin network than CytoD. The actin network became less organized, but individual fibers could still be observed. Figure 3.1: Immunofluorescent stainings visualizing actin fibers (red) and cell nuclei (blue) in control samples (A,B) and samples treated with 50% FBS (C,D), CytoD (E,F) and ROCK inhibitor (G,H). Control and FBS samples show aligned and organized fiber structures. Cells treated with CytoD become rounded at the surface layer and show a fragmented network at 28 µm. ROCK inhibitor causes a slight disruption of the actin fiber organization. The white bar represents 50 µm. Figure 3.2 and table 3.1 provide an overview of the force measured before (pre) and after (post) incubation with the biochemicals, and the absolute and relative change. On average, generated force after 4 weeks of culture was 39.4 ± 2.3 mn. Incubation with FBS for 2 hours had no effect on the generated force within the TE constructs, while incubation with CytoD and ROCK inhibitor significantly decreased the generated force by 39

48 Chapter ± 0.7 mn and 6.8 ± 0.5 mn, respectively. There was no difference in the effect of treatment between CytoD or ROCK inhibitor, both reduced force by approximately 20%. DNA content was similar in all groups, so cell density could not have caused differences in the generated force or changes to it (data not shown). Figure 3.2: Generated force in control samples (A) and before (pre) and after (post) incubation with 50% FBS (B), CytoD (C) and ROCK inhibitor (D). (E) Change in force relative to the generated force before addition of the biochemicals. Force remained constant in samples treated with 50% FBS and decreased in samples treated with CytoD and ROCK inhibitor. ** denotes a significant difference of p<0.001 to control samples. 40

49 Passive and active contributions to generated force and retraction Retraction after release of the constraints occurred rapidly during the first 30 minutes, after which it slowly continued (Figure 3.3A). After 2 hours, retraction was higher in the FBS group than in the control samples. Incubation with CytoD and ROCK inhibitor decreased the amount of retraction compared to control samples. There was no significant difference between incubation with ROCK inhibitor or CytoD (Figure 3.3B). An overview of the absolute retraction values is given in table 3.1. Figure 3.3: (A) Tissue retraction during 2 hours after release of constraints. (B) After 2 hours, retraction was higher in the FBS samples and lower in CytoD and ROCK inhibitor samples compared to the control samples. * denotes a significant difference of p<0.05 to control samples. Retraction of the FBS samples remained higher up to 10 hours, but after 24 hours retraction in control samples reached a similar level (Figure 3.4). Two hours after refreshment of the medium, the ROCK inhibitor and CytoD samples started to recover. Recovery seemed faster in the ROCK inhibitor samples than in the CytoD samples. After 8 hours, retraction was no longer significantly lower than in the control samples. Table 3.1: absolute and relative force and retraction. Force change Retraction [%] Absolute [mn] Relative [%] 2h % control 2h 10h 24h 48h Control 1.1 ± ± ± ± ± ± ± 6.6 FBS 2h ± ± ± 2.4 * ± 10.0 * 49.5 ± 1.1 * 51.7 ± ± 1.9 CytoD 2h 7.7 ± 0.7 ** 20.1 ± 1.9 ** 14.8 ± 1.3 * 61.0 ± 5.4 * 27.9 ± ± ± 5.5 ROCK 2h 6.8 ± 0.5 ** 21.0 ± 3.2 ** 14.7 ± 1.0 * 60.7 ± 4.0 * 31.7 ± ± ± 4.9 Control ± ± ± ± ± ± 2.7 Lysed 12.9 ± 1.9 ** 27.3 ± 3.4 ** 4.2 ± 0.7 ** 17.5 ± 2.7 ** 4.3 ± 0.8 ** 5.1 ± 0.9 ** Decellularized 9.5 ± 1.3 * 25.4 ± 2.7 * 3.7 ± 0.3 ** 15.2 ± 1.3 ** 5.9 ± 0.6 ** 5.9 ± 0.5 ** Table 3.1: * and ** denote a significant difference of respectively p<0.05 and p<0.001 compared to control samples. 41

50 Chapter 3 Figure 3.4: (A) Tissue retraction and recovery from biochemicals over time. (B) After 8 hours of recovery, retraction of CytoD and ROCK inhibitor samples was similar to control samples, while retraction of FBS samples was higher than control samples. At 24 and 48 hours retraction was similar in all samples. * denotes a significant difference of p< 0.05 to control samples The effect of cell lysis and decellularization on force and retraction In the lysed constructs the amount of DNA was similar to the control samples (Figure 3.5A). The DAPI and HE stainings confirmed the presence of cells (Figure 3.5E,H). Decellularized constructs contained hardly any detectable DNA, which was confirmed in the staining by the absence of cell nuclei in the DAPI staining and cells in the HE staining (Figure 3.5A,F,I). The amount of collagen remained unaffected by both lysis and decellularization (Figure 3.5B). There was a significant decrease in GAGs of both lysed and decellularized constructs compared to controls (Figure 3.5C). This decrease was larger in the decellularized samples than in the lysed samples. 42

51 Passive and active contributions to generated force and retraction Figure 3.5: DNA (A), collagen (B) and GAG (c) content, and DAPI (D F) and HE (G I) staining of control, lysed and decellularized TE constructs. In the control and lysed samples, cells were present (small, dark dots indicated by the white arrows). In the decellularized constructs cells were absent. GAG content decreased in lysed and decellularized samples. The black arrows indicate scaffold remnants. ** denotes a significant difference of p<0.001 to control samples. # denotes a significant difference of p<0.05. The white bars represent 200 µm. Lysis and decellularization decreased the generated force by 12.9 ± 1.9 mn and 9.8 ± 1.2 mn, respectively (Figure 3.6, Table 3.1). There was no difference between the effects of lysis and decellularization on the generated force. The relative decreases compared to the force before lysis and decellularization were respectively 27.3 ± 3.4% and 25.4 ± 2.7%. 43

52 Chapter 3 Figure 3.6: Generated force in control samples (A) and before (pre) and after (post) lysis (B) and decellularization (C). (D) Change in force relative to the generated force before addition lysis and decellularization. Force decreased in the lysed and decellularized samples compared to the control samples. ** denotes a significant difference of p<0.001 to control samples. At release of constraints, retraction in the lysed and decellularized samples was substantially lower than in the control samples at 2, 10 and 24 hours (Figure 3.7). Retraction in the lysed and decellularized samples occurred in the first 30 minutes, after which they remained stable, while control samples kept on retracting (Figure 3.7). An overview of the absolute retraction values is given in table

53 Passive and active contributions to generated force and retraction Figure 3.7: (A) Tissue retraction during 24 hours after release of constraints. (B) Retraction after 2, 10 and 24 hours was less in the lysed and decellularized samples compared to control samples. ** denotes a significant difference of p<0.001 to control samples. 3.4 Discussion Cell mediated compaction and retraction are general occurring phenomena in tissue engineering, mainly observed and investigated in fibrin and collagen gel systems. Within heart valve tissue engineering, cell traction forces lead to leaflet shrinkage, which, in vivo, results in regurgitation. The difference between an autologous tissue engineered heart valve and the generally investigated gel systems is the presence of a biodegradable synthetic scaffold and a well developed ECM that is synthesized by the cells. As these two aspects complicate tissue retraction, thorough investigation of the problem is needed, before it can be solved. Previously, we have quantified the generation of force, stress, compaction and retraction over time, and observed that a 45

54 Chapter 3 fully developed tissue compacts and retracts significantly, but less than the gel systems do (Van Vlimmeren et al., 2011a). In this study, we aimed to quantify the contribution of the cells, as retraction is a cell mediated problem. Therefore, we distinguished passive from active contributions to generated force and retraction, by changing traction forces Passive and active components in force generation Increasing the FBS concentration from 10% to 50% did not affect the generated force. It was predicted that increasing the FBS concentration would increase the generated force, as this was shown when increasing the FBS concentration from 0 10% to 20% (Wakatsuki et al., 2000; Yee et al., 2001). FBS contains many growth factors, from which transforming growth factor β (TGF β) is one that is known to induce cell traction forces (Brown et al., 2002; Hinz, 2006; Wipff and Hinz, 2009). Although TGF β probably is not the only growth factor in FBS that induces increased traction forces, for this particular growth factor it has been shown that there is an optimum working concentration. Concentrations above this optimum, decreased the generated force (Brown et al., 2002). 50% FBS could be above this optimum, but as it did increase tissue retraction, as discussed later, this is unlikely. Increased cellular traction forces might be counteracted by the developed ECM, while in tissue retraction the pre stress in the ECM is released, which makes it vulnerable to the traction forces of the cells. Incubation with CytoD or ROCK inhibitor decreased the generated force by ~20% compared to control samples. Although CytoD disrupts the complete actin network and ROCK inhibitor only inhibits myosin actin contraction, there was no difference between the two treatments in generated force. The actin fiber network of the CytoD treated samples was fragmented compared to the ROCK inhibitor samples, where individual fibers were still present. To validate the elimination of cellular traction forces, we incubated TE constructs with CytoD for 16 hours, but this did not change the decrease in force (data not shown). These results suggest that it is irrelevant how the cellular traction forces are eliminated. Next, we lysed and decellularized the TE constructs to eliminate potential remaining passive cell contributions. The resulting decrease in force was slightly more than in CytoD and ROCK inhibitor samples (~26% opposed to ~20%), but this was not significant. In literature, the reported active and passive cellular contributions to generated force are much higher. The active cellular contribution to force of fibroblasts in collagen gels was 35 65% (Legant et al., 2009; Marquez et al., 2009; John et al., 2010; Youssef et al., 2011) and lysis caused an additional 30% reduction of force below the level observed with CytoD treatment (Wakatsuki et al., 2000). To measure a decrease in force within our model system, bending of the leaf springs must become less and this requires elongation of the TE constructs. As cells have remodeled the ECM to a certain length, relaxation of the ECM can probably occur up to a maximum. The 20 25% decrease in force that we obtained, is closest to the 46

55 Passive and active contributions to generated force and retraction observation of John et al., who observed a decrease of ~35% after incubation with CytoD in a set up that also used bending beams Passive and active components in tissue retraction Two hours after release of constraints, retraction in the FBS samples was ~45% higher than in the control samples. CytoD or ROCK inhibitor decreased retraction by ~40% compared to control samples, and lysis and decellularisation of the samples decreased retraction by ~85%. We have summarized these results in Figure 3.8, illustrating activated retraction and a subdivision of normal retraction into active and passive components. Figure 3.8: Illustration of active and passive cell retraction, passive ECM retraction and activated cell retraction in TE constructs. Active cell retraction accounts for ~40% of total retraction, passive cell retraction accounts for ~45% of total retraction and passive residual stress in the ECM accounts for ~15%. Finally, activated cells increase retraction by 45%. FBS increased tissue retraction by 45%. The growth factor TGF β, which we previously indicated to regulate traction forces of the cells has also shown to increase retraction (Grouf et al., 2007). As in vivo plasma levels reach 55% of serum, this increased retraction is important for the in vivo response of the cells after implantation. The high plasma level in vivo is likely to enhance cellular traction forces and retraction. However, the final retraction level will not change as retraction of control samples reached similar levels as the FBS samples after 24 hours in vitro, most likely caused by volume restrictions. 47

56 Chapter 3 In fibroblast seeded fibrin gels, active retraction accounted for ~75% of the total retraction (Grouf et al., 2007; Balestrini and Billiar, 2009) and in collagen gels seeded with fibroblasts, this was ~80% (Tomasek et al., 1992). Based on this literature, the active retraction in the TE constructs was less than expected. 16 hours of incubation with CytoD, to check the elimination of traction forces, did not change retraction compared to 2 hours of incubation (data not shown). Therefore, we eliminated passive cellular contributions as well by lysis and decellularization. Although passive cell behavior has been reported by others (Wakatsuki et al., 2000; Roy et al., 2009), it was not expected to contribute for 45% to retraction. In porcine arteries, elongation increased from 12% to 20% between CytoD treatment and decellularization (Roy et al., 2009). Passive contribution might arise from cellular cross linking of the matrix or the space occupying properties of the cells (Wakatsuki et al., 2000). However, both require that the cells remain elongated and connected to the ECM. For both ROCK inhibition and CytoD treatment, it was indeed shown that cells become irregular in shape, but retain their elongated cell configuration in this and other studies (Tomasek and Hay, 1984; Tamariz and Grinnell, 2002). The absence of differences between lysed and decellularized samples demonstrated that the presence of cellular material does not affect retraction once cells are lysed. The residual stress within the ECM accounted for 15% of total retraction. This residual stress is likely developed during ECM remodeling of the cells, in which collagen fibers are re organized and cross links are formed. Cell lysis and decellularization decreased the amount of GAGs. As, we have previously shown that increased GAG content decreased the amount of retraction (van Vlimmeren et al., 2011b), the passive retraction of the ECM could have been overestimated in this study and partly be caused by the decrease in GAGs Implications for heart valve tissue engineering It has been hypothesized that temporarily decreasing cellular traction forces at time of implantation might solve the problem of retraction. This hypothesis is based on the idea that loads applied during the cardiac cycle might be able to counterbalance the traction forces once the valve is functioning in vivo. The results presented in this study demonstrate that this hypothesis does not hold. When eliminating the traction forces of the cells the TE constructs still retracted up to 15% within 2 hours. Furthermore, maximum loading only occurs a very brief period every heart cycle, and further studies are needed to investigate whether this loading is enough to resist the continuous presence of the traction forces. As soon as the valve is implanted, cellular traction forces will recover. Recovery from CytoD and ROCK at 20% FBS occurred within 8 hours and in vivo this will likely be faster due to higher serum concentrations. 48

57 Passive and active contributions to generated force and retraction Decellularization and lysis reduced retraction to ~4%. Although this is still a lot, it is easier to correct for in the design of the heart valve as retraction occurs immediately upon release of constraints and remains stable after that. Implantation of a decellularized valve would represent a different approach, as it would require seeding of new non contractile cells or the attraction of endogenous cells in vivo to ensure a viable implant. It is, however, a promising approach as decellularized valves have already shown functionality in vivo (Simon et al., 2003; Cebotari et al., 2006; Erdbrugger et al., 2006; Hopkins et al., 2009) and it is currently under investigation for tissue engineered constructs as well (Dahl et al., 2011) Conclusion The active cellular contribution to the generated force after 4 weeks is at least 20%. Passive retraction accounted for 60% of total retraction, of which 15% was residual stress in the ECM and 45% was passive cell retraction. Cellular traction forces account for the remainder of the retraction (40%). These results indicate that the passive cell contribution to retraction should not be underestimated. Interpretation of passive cell retraction is difficult, as the effect of the biochemicals on the interaction between the cell and the matrix is uncertain. Ideally, one would like to bypass the passive cell contribution, without lysis of the cells to maintain a viable implant. The residual stress in the ECM is small and might be overcome by design changes to the tissue engineered heart valve geometry. These results provide valuable insights into the passive and active components of tissue retraction and are crucial for solving leaflet retraction in heart valve tissue engineering. 3.5 Acknowledgements The authors gratefully acknowledge the support of the Smart Mix Program of the Netherlands Ministry of Economic Affairs and the Netherlands Ministry of Education, Culture and Science. The authors would like to thank Nicky de Jonge for the visualization of the actin staining and Linda Kock for performing the biochemical assays. 49

58 Chapter 3 50

59 Chapter 4 Controlling matrix formation and cross linking by hypoxia in cardiovascular tissue engineering This chapter is based on: Marijke A.A. van Vlimmeren, Anita Driessen Mol, Marloes W.J.T. van den Broek, Carlijn V.C. Bouten, Frank P.T. Baaijens (2010) Controlling matrix formation and cross linking by hypoxia in cardiovascular tissue engineering J. Appl. Physiol., 109( ),

60 Chapter Introduction Cardiovascular diseases, such as heart valve dysfunction and coronary heart disease, represent a major health problem to the population. Worldwide approximately heart valve replacement surgeries are performed annually (Pibarot and Dumesnil, 2009), while adults undergo a bypass procedure in the USA alone (Writing Group et al.). Current available heart valve replacements are either mechanical or bioprosthetic in nature. In the case of coronary or peripheral artery diseases, a bypass of an autologous or synthetic vessel is needed. These replacement valves and synthetic vessels are unable to fully restore native valve behavior, because they are made of nonliving material lacking the capability to adapt to changing environmental hemodynamics by continuous remodeling. Cardiovascular tissue engineering is considered a promising technique to overcome this problem. A living autologous tissue engineered valve or vessel would be able to grow, repair and remodel during life, which is especially important for pediatric and young adult patients. Tissue engineered constructs are generally composed of a biodegradable scaffold seeded with autologous human myofibroblasts that produce extracellular matrix. A main challenge in cardiovascular tissue engineering is the in vitro production of an extracellular matrix mimicking native structural composition and hence appropriate functional tissue behavior when subjected to in vivo physiological loads. The extracellular matrix of native cardiovascular tissue is composed of proteoglycans and fibrillar proteins, such as elastin and collagen. The elastin fibers are mainly important for the elastic behavior of the tissue, while the collagen fibers (mainly types I and III) form the most prominent load bearing component. In addition to the collagen content and organization, the load bearing capacity of collagen is highly dependent on collagen cross linking, which stabilizes the collagen fibrils (Balguid et al., 2007). The in vitro formation of these highly important extracellular matrix components can be affected by mechanical, biochemical (e.g. vitamins) and environmental stimuli, such as oxygen concentration (Mol et al., 2005; Balguid et al., 2009; Rubbens et al., 2009). Detailed knowledge of the individual effects of these stimuli is important, because it allows for the in vitro regulation of tissue formation. Cell culture studies are generally performed at 21% O 2, which should be considered as hyperoxia when compared to physiological oxygen levels in blood ranging from 5% O 2 in venous blood to 13% O 2 in arterial blood (Guyton and Hall, 2000). Oxygen levels below these physiological concentrations are termed hypoxia and have a strong impact on cell biology. The process behind hypoxia is not completely known yet, but the transcription factor hypoxia inducible factor 1 (HIF 1) is believed to be one of the key regulators responsible for the induction of genes during hypoxic conditions (Ke and Costa, 2006; Pak et al., 2007). At present, there are more than 100 HIF 1 activated genes identified with varying functions (Ke and Costa, 2006). Several studies in 2D and 3D cell cultures have been conducted to investigate the influence of hypoxia on collagen 52

61 Controlling matrix formation and cross linking by hypoxia (Falanga et al., 1993; Agocha et al., 1997; Horino et al., 2002; Hofbauer et al., 2003; Brinckmann et al., 2005), and collagen cross link formation (Hofbauer et al., 2003; Brinckmann et al., 2005). For example, hypoxia has been associated with an increase of the collagen cross link enzyme Lysyl Hydroxylase 2 (LH2), but procollagen α1(i) changes due to hypoxia differed depending on the cell source used (Agocha et al., 1997; Horino et al., 2002; Hofbauer et al., 2003; Brinckmann et al., 2005). In addition, in vitro findings varied depending on the oxygen concentration to which the cells were exposed and the time of exposure. In vivo studies showed increased mrna levels of procollagen α1(i) and α1(iii) in mice exposed to hypoxia and rat hearts with an induced myocardial infarct (Takeda et al., 2000; Estrada and Chesler, 2009). Indeed, hypoxia may represent a promising stimulus to optimize tissue formation for cardiovascular tissue engineering. Therefore, in the present study the effect of hypoxia is investigated on human vascularderived myofibroblasts, a cell source often used in cardiovascular tissue engineering because of their matrix producing capacity (Mol et al., 2006; Stekelenburg et al., 2009). Recently, it was demonstrated that human vascular derived myofibroblasts seeded onto a biodegradable scaffold and cultured at the physiological oxygen concentration of 7% gave an increase in the total amount of collagen and collagen crosslinks compared to constructs cultured at 21% of oxygen (Balguid et al., 2009). The effect of oxygen concentrations below 7% on extracellular matrix production of human myofibroblasts has still to be investigated. As it is known that oxygen concentrations at the cellular level need to be below 5% to induce significant HIF 1 stabilization (Jiang et al., 1996; Brooks et al., 2009) and that HIF 1 induces upregulation of various matrix regulated genes (Ke and Costa, 2006), we hypothesize that decreasing the oxygen concentration below 7% might even further improve tissue formation. 3D tissue engineered constructs have an oxygen gradient within the construct due to poor oxygen diffusion into the centre of the tissue (Brown et al., 2007). This impairs the investigation of the differential effects of individual oxygen concentrations. Therefore, in the present study the effect of decreasing oxygen concentrations on the hypoxic state and extracellular matrix production is studied in monolayers of human myofibroblasts. The hypoxic state of cells is represented by the expression of HIF 1. However, HIF 1 is difficult to measure. It consists of HIF 1α and HIF 1β, from which the protein HIF 1α is very unstable at 21% O 2, while its mrna is not affected by low oxygen concentrations (Ke and Costa, 2006). Therefore, the more stable vascular endothelial growth factor (VEGF), a well known HIF 1 target gene, is used to indicate the hypoxic state of the cells (Shima et al., 1995; Forsythe et al., 1996). Changes in the extracellular matrix production are measured at gene expression and protein level over a range of decreasing oxygen concentration from 7 to 0.5%. In this study 21% O 2 is defined as hyperoxia, 7% O 2 as normoxia and 4 to 0.5% O 2 as hypoxia. Knowing the optimal oxygen concentration for human myofibroblasts to produce extracellular matrix components offers the potential to regulate tissue formation, as required to optimize tissue properties. 53

62 Chapter Materials & Methods Myofibroblast cell culture Human myofibroblasts were harvested from the vena saphena magna, obtained from a 63 years old female patient according to the Code for proper secondary use of human tissue in the Netherlands as provided by the Dutch Federation of Medical Scientific Societies (FMWV). These cells have previously been characterized as myofibroblasts (Mol et al., 2006) showing the expression of vimentin, but not desmin. A subpopulation of the cells expressed α smooth muscle cells. Cells were expanded in T75 flasks until passage 5 in a humidified atmosphere containing 5% CO 2 at 37 C using standard cell culture methods (Schnell et al., 2001). The culture medium consisted of advanced Dulbecco s Modified Eagle Medium (DMEM; Invitrogen, Carlsbad, USA), supplemented with 10% Fetal Bovine Serum (FBS; Greiner Bio one, Frickenhausen, The Netherlands), 1% GlutaMax (Invitrogen), and 1 % penicillin/streptomycin (Lonza, Basel, Switzerland) Experimental design Monolayers of cells (3000 cells/cm 2, passage 5) were cultured for at least 3 days at 21% O 2 (hyperoxia). At 50 70% confluency, medium was refreshed and flasks were exposed to hyperoxia, normoxia or hypoxia in a CO 2 /O 2 controllable incubator (MCO 18M SANYO; Wood Dale, USA, 5% CO 2, 37 C). Within the gene expression measurements, for every flask placed at a low oxygen condition, a flask was placed at hyperoxia to serve as a control. Exposure time and oxygen concentrations were chosen based on pilot experiments. After 24 hours of exposure to 7, 4, 2, 1 and 0.5% O 2, cell viability and morphology was checked. Additionally, gene expression measurements were performed to measure changes of relevant extracellular matrix components and cross link enzymes compared to the control cultures at hyperoxia (n=9 per oxygen concentration). Based on the results of these gene expression measurements, additional experiments were performed in which cell viability and morphology, proliferation rates, and protein synthesis were investigated at oxygen concentrations of 7, 2 and 0.5% O 2 after 4 days of exposure and compared to hyperoxia (n=3 6 per oxygen concentration). Protein synthesis was evaluated with immunocytochemistry (n=3), an enzyme immunoassay to assess collagen synthesis (n=6), and Western blotting to assess the cross link enzyme LH2 (n=3). All changes at gene expression and protein level, from both normoxic (7% O 2 ) and hypoxic (4 0.5% O 2 ) conditions, have been compared to control hyperoxic conditions (21% O 2 ). A schematic overview of the experimental design is provided in figure

63 Controlling matrix formation and cross linking by hypoxia Figure 4.1: Schematic overview of the experimental design Cell morphology, viability and proliferation Cell morphology was analyzed from microscopic phase light contrast images. From these images the aspect ratios of 30 cells per oxygen concentration were estimated using ImageJ (WCIF ImageJ, National Institutes of Health, USA). To examine cell viability, cells were cultured on glass coverslips (Ø 13 mm) coated with 0.1% gelatin. After exposure to hyperoxia, normoxia or hypoxia, samples were immediately analyzed for cell viability with a live/dead staining. Cells were stained with Cell Tracker Green (CTG; Invitrogen), which stains living cells, and Propidium Iodide (PI; Invitrogen), which binds to the DNA of dead cells. Cells were incubated with medium containing 10 µm CTG at 37 C for 20 minutes. Subsequently, cells were washed with phosphate buffered saline (PBS; Sigma Aldrich, St. Louis, USA) and incubated with medium containing 7 µm PI for 30 minutes at 37 C. Samples were visualized with fluorescence microscopy (Axiovert 200M, Zeiss, Göttingen, Germany). Cell proliferation was measured during 4 days of exposure to hyperoxia, normoxia or hypoxia. Cells were seeded in 6 wells plates (2000 cells/cm 2 ) and cultured at hyperoxia for 3 days. Subsequently, they were placed at the desired oxygen 55

64 Chapter 4 concentration (day 0). During the following 4 days, 3 wells were sacrificed every day and the number of cells was counted with a NucleoCounter (Chemometec, Allerød, Denmark) Gene expression Gene expression levels of vascular endothelial growth factor (VEGF), procollagen α1(i) (col I) and α1(iii) (col III), elastin, and cross link enzymes lysyl oxidase (LOX) and lysyl hydroxylase 2 (LH2) were measured. After exposure to either hyperoxia, normoxia or hypoxia, cells were washed twice with PBS and immediately lysed by scraping with RLT buffer to minimize RNA degradation. Total RNA was isolated using the Qiagen Micro Kit according to the manufacturer s instructions (Qaigen, Hilden, Germany). Nucleic acid content was determined spectrophotometrically (NanoDrop ND1000, Isogen Life Science, IJsselstein, The Netherlands). Synthesis of cdna was carried out with 1000 ng of RNA in a 50 µl reaction volume consisting of 0.5 mm dntps (Invitrogen), 2 µg/ml random primers (Promega, Madison, USA), 10 mm DTT (Invitrogen), 4 IU/µl M MLV (Invitrogen), M MLV buffer (Invitrogen), and ddh2o. Control reactions without M MLV ( RT) were performed to screen for genomic DNA contamination. The temperature profile of the cdna synthesis protocol was 6 minutes at 72 C, 5 minutes at 37 C (with subsequent addition of M MLV), 60 minutes at 37 C, and 5 minutes at 95 C. Samples were stored at 4 C for qpcr use. Stability of reference genes with respect to the experimental treatment was determined by the genorm algorithm described by Vandesompele et al. (Vandesompele et al., 2002). The two most stable reference genes were cytochrome c 1 (CYC1) and β 2 microglobulin (B2M). Primer sets of the target genes were developed or taken from literature and validated for qpcr (Table 1). The qpcr reaction (MyiQ, Bio Rad, Hercules, USA) was performed using SYBR Green Supermix (Bio Rad), primers (0.5 mm), ddh2o, and 1 µl undiluted cdna in a 15 µl reaction volume. The temperature profile was 3 minutes 95 C, 40x (20s 95 C, 20s 60 C, 30s 72 C), 1 minute 95 C, 1 minute 65 C, followed by a melting curve analysis Western blot analysis Pellets of cells were incubated with 10 µl of RIPA buffer (Sigma Aldrich) and 10 µl Laemmli buffer at 100 C for 5 minutes to denaturate the proteins. Proteins were separated by SDS PAGE (Mini PROTEAN 3 System, Bio Rad) on a 10% poly acrylamide gel and transferred to a PVDF membrane (Millipore Corporate, Billerica, MA, USA). The membrane was blocked by incubation in PBS/0.1% Tween 20/5% milk solution for 1 hour and probed overnight at 4 C with either mouse anti human LH2 or mouse antihuman beta actin (both derived from Abcam, Cambridge, UK) in PBS/0.1% Tween 20/1% 56

65 Controlling matrix formation and cross linking by hypoxia milk. After washing and incubation with the conjugate rabbit anti mouse peroxidase (Pierce, Rockford, IL, USA) in PBS/0.1% Tween 20 for 1 hour the blots were developed using Supersignal West Dura chemiluminescent Substrate (Pierce) and visualized by Versadoc (Bio Rad) Immunocytochemistry Immunofluorescent stainings were performed on cells cultured on glass coverslips (Ø 13mm) coated with 0.1% gelatin. Cells were washed with PBS (Sigma Aldrich), fixated with 3.7% formalin (Sigma Aldrich) for 20 minutes and permeabilized with 0.5% Triton X 100 (Merck, Schiphol rijk, The Netherlands) in PBS for 10 minutes. Non specific binding was blocked by incubating twice with 1% horse serum (HS, Invitrogen) in PBS for 10 minutes, after which cells were washed 2 times for 10 minutes with NET gel (50mM Tris ph 7.4, 150mM NaCl, 5mM EDTA, 0.05% NP40, 0.25% gelatin). Subsequently, cells were incubated for 2 hours with primary antibodies against collagen type I (1:100) (Sigma Aldrich), collagen type III (1:200) (Abcam) and LH2 (1:200) (Abcam) in NET gel with 10% HS. Next, cells were washed 6 times with NET gel and incubated for 30 minutes with Alexa fluor 488 conjugated goat anti mouse IgG1 (1:300) secondary antibody for collagen type I and LH2, and Alexa fluor 488 conjugated goat anti rabbit IgG(H+L) secondary antibody for collagen type III (A21121 and A11008, Invitrogen). Subsequently, cells were washed 3 times with NET gel and stained for an additional 5 minutes with 4, 6 diamidino 2 phenylindole (DAPI, Sigma Aldrich). Finally, cells were washed 4 times with PBS and mounted on slides with mowiol (Calbiochem, San Diego, USA). Cells were evaluated using fluorescence microscopy (Axiovert 200M, Zeiss, Göttingen, Germany). Immunofluorescent intensity of the stainings has been quantified using Matlab (The MathWorks, Eindhoven, The Netherlands) and normalized to cell number and control cultures at hyperoxia Procollagen Type I C peptide quantification Collagen synthesis has been measured indirectly through the amount of procollagen type I carboxy terminal peptide (PIP) in the culture supernatents. The concentration of PIP was measured with an in vitro enzyme immunoassay (EIA) kit (TAKARA BIO INC.,Otsu, Japan) according to the manufacturer s instructions. To review briefly, 100 µl of antibody peroxidase conjugate solution (provided in EIA kit) and 20 µl of diluted sample were transferred to each well and incubated at 37 C for 3 hours. Subsequently, the wells were washed 4 times with PBS. Next, 100 µl of substrate solution (provided in EIA kit) was added to each well and incubated at room temperature for 15 minutes. Finally, 100 µl stop solution (1N H 2 SO 4 ) was added and absorbance of samples and standards were measured in duplicate at 450 nm with a 57

66 Chapter 4 plate reader (Synergy HT microplate reader, Winooski, USA, Bio Tek). The measured PIP concentration was normalized to cell number and control cultures at hyperoxia Statistical analysis All data are presented as mean and their standard deviations. Relative expression levels were normalized to the geometric mean of both reference genes and the control group at hyperoxia (Hellemans et al., 2007; Bustin et al., 2009). Differences between sample and control (hyperoxia) were assessed statistically with an unpaired student s t test for gene expression levels, as measuring the effect of each individual low oxygen concentration compared to hyperoxia was an experiment in itself. For the aspect ratio, collagen synthesis and intensity of the immunofluorescent staining a one way ANOVA with a Tukey s multiple comparison post hoc test was used. Differences were analyzed using GraphPad Prism software (GraphPad Software, Inc, USA) and considered significant for P values < Results Morphological integrity, cell viability and proliferation After 24 hours, cell shape and size were comparable for all oxygen concentrations. After 4 days, cells exposed to 7% and 2% O 2 also retained both normal morphology (Figure 4.2A C,I) and proliferation rates (Figure 4.2J) comparable to the cells exposed to hyperoxia. However, cells exposed to 0.5% O 2 for 4 days, tended to lose their spindle like shape and became more stretched, as reflected by a decreased aspect ratio (Figure 4.2D,I). Further, proliferation at 0.5% O 2 decreased after 4 days compared to cells cultured at hyperoxia (Figure 4.2J). Cells retained viability for all oxygen concentrations after both 24 hours and 4 days (Figure 4.2E H). No changes in viability were observed among the groups Gene expression At hyperoxia, procollagen α1(i), α1(iii) and elastin were all expressed, particularly procollagen α1(i) which was approximately 10 times higher than expression levels of procollagen α1(iii) and 40 times higher than those of elastin (Figure 4.3A). Furthermore, expression levels of the cross link enzymes differed at hyperoxia, with LOX being 20 times higher than those of LH2 (Figure 4.3B). 58

67 Controlling matrix formation and cross linking by hypoxia Figure 4.2: (A D) Phase contrast light microscopy images: human myofibroblasts exposed to 21%, 7%, 2% O 2 for 4 days retain their normal morphology (A C), while myofibroblasts exposed to 0.5% O 2 lose their spindle like shape (D). (E H) Myofibroblasts remain viable after exposure to 21%, 7%, 2% and 0.5% O 2 for 4 days (PI/CTG staining; green: viable cells, CTG; red: dead cells). (I J) Aspect ratio and proliferation rate of myofibroblasts decreases after 4 days of exposure to 0.5% O 2. Myofibroblasts exposed to 7% and 2% O 2 retain a normal aspect ratio and proliferation rate. * and ** denote a significant difference of respectively p<0.05 and p<0.001 compared to controls cultured at hyperoxia. Figure 4.3: (A) Gene expression levels of procollagen 1α (I) were 10 times higher than expression levels of procollagen 1α (III), and 40 times higher than those of elastin in human myofibroblasts at hyperoxia. (B) Gene expression levels of LOX were 20 times higher than those of LH2 (B) in human myofibroblasts at hyperoxia. 59

68 Chapter 4 Figure 4.4: (A C) Gene expression levels of VEGF (A); procollagen 1α (I) and (III), and elastin (B); and LH2 and LOX (C); in human myofibroblasts after exposure to various oxygen concentrations for 24 hours. VEGF and LH2 gene expression levels increase at oxygen concentrations of 4% and lower. LOX, and procollagen 1α (I) and (III) gene expression levels increase at oxygen concentrations of 2% and lower. Elastin gene expression levels remain unaffected. * and ** denote a significant difference of respectively p<0.05 and p<0.001 compared to controls cultured at hyperoxia. Gene expression levels are normalized to the control group at hyperoxia (dotted line). (D E) Exponential fits to the data points (solid squares, 9 per oxygen concentration) of VEGF (D), LH2 (E) and LOX (F). All show an exponentially increased gene expression with decreasing oxygen concentration. The dotted lines indicate the 95% confidence intervals. 60

69 Controlling matrix formation and cross linking by hypoxia VEGF gene expression levels increased exponentially at 4% O 2 and lower (Figure 4.4A,D). Therefore, we defined oxygen concentrations of 4% and lower as hypoxia. Gene expression levels of procollagen α1(i) and α1(iii) increased to 138±26% and 143±19% (p<0.05) respectively, for all oxygen concentrations below 4% compared to hyperoxia (Figure 4.4B). By contrast, elastin gene expression levels remained unaffected with decreasing oxygen concentrations (Figure 4.4B). The largest effects were seen for the cross link enzymes LH2 and LOX. Gene expression levels of both enzymes increased exponentially with decreasing oxygen concentrations, with smaller effect evident for LOX than for LH2 (228 ± 45% versus 611 ± 176% at 0.5% O 2 compared to hyperoxia) (Figure 4.4C,E,F). Culturing at 7% O 2 had no effect on the gene expression levels for any of the examined genes compared to hyperoxia Protein synthesis The amount of PIP in the supernatant after 4 days of culture increased at oxygen concentrations of 2% and 0.5% compared to hyperoxia (Figure 4.5A). Immunocytochemistry showed that at hyperoxia all cells stained positively for collagen type III, while only a proportion of cells stained positively for collagen type I (Figure 4.6A,E). No difference was observed in intensity of collagen type I staining at 7%, 2% and 0.5% O 2 compared to hyperoxia (Figure 4.6A D,M). Collagen type III staining was more pronounced at hyperoxia, 7% and 2% than at 0.5% O 2 (Figure 4.6E H,N). Figure 4.5: (A) Collagen synthesis, normalized to cell number and control cultures at hyperoxia, was measured through PIP analyses in the supernatant of human myofibroblasts. Collagen synthesis increases when cells are exposed to 2% or 0.5% O2 for 4 days and is not affected by culturing at 7% O2. * and ** denote a significant difference of respectively p<0.01 and p<0.001 compared to controls cultured at hyperoxia (dotted line). (B) The amount of LH2 produced by human myofibroblasts increases when cells are exposed to 2% and 0.5% O2 for 4 days. LH2 levels at 7% O2 remain unaffected. 61

70 Chapter 4 With respect to production of the cross link enzyme LH2, the exposure of cells to 2% and 0.5% O 2 for 4 days produced a significant increase in the amount of LH2 compared to cells at hyperoxia (Figure 4.5B; Figure 4.6I L,O). No differences were observed between cells at 7% O 2 and hyperoxia (Figure 4.5B; Figure 4.6O). Figure 4.6: (A H) Representative examples of collagen type I (A D), type (III) (E H) and LH2 (I L) staining (green) of human myofibroblasts exposed to 21% (A, E, I), 2% (B, F, J), 7% (C, G, K) and 0.5% O 2 (D, H, L). Cell nuclei were stained with DAPI (blue). (M O) Quantified intensity of immunofluorescent stainings normalized to cell number and control cultures at hyperoxia. LH2 staining is most pronounced at 2% and 0.5% compared to 7% and 21% O 2, while collagen type III is more pronounced at 21%, 7% and 2% than at 0.5% O 2. No difference was observed in intensity of collagen type I staining at 7%, 2% and 0.5% O 2 compared to hyperoxia. * and ** denote a significant difference of respectively p<0.01 and p<0.001 compared to controls cultured at hyperoxia (dotted line). 62

71 Controlling matrix formation and cross linking by hypoxia 4.4 Discussion Optimization of cardiovascular tissue engineering for human implantation requires in vitro regulation of tissue properties. It is a challenge to provide an environment that stimulates matrix production and maturation, thereby improving tissue properties to meet in vivo functional demands. This study elaborated on the effects of the environmental factor oxygen concentration on extracellular matrix production. The effects of low oxygen concentrations are widely investigated for application in bone and cartilage tissue engineering (Volkmer et al., 2008; Das et al., 2009), stem cell culture (Ma et al., 2009) and its role in tumors and cardiac infarcts (Takeda et al., 2000; Rankin and Giaccia, 2008). However, results vary between cell sources and for human myofibroblasts, a commonly used cell source in cardiovascular tissue engineering, the effects have not been established. Therefore, we investigated changes in growth, viability and ECM production of these cells due to culturing at low oxygen concentrations. In our study, we cultured human myofibroblasts at hyperoxia (21% O 2 ), normoxia (7% O 2 ) and hypoxia (4 0.5% O 2 ). We demonstrated that culturing at normoxia has no effect on any of the investigated genes and proteins compared to culturing at hyperoxia. Human myofibroblasts convert to hypoxia driven gene expression, represented by VEGF expression, when exposed to oxygen levels of 4% or lower. As matrix genes are affected by hypoxia, these results imply that to obtain effects of culturing at low oxygen concentrations on the extracellular matrix level, the oxygen concentration at the cellular level has to be 4% or below. The exponential increase of hypoxia driven gene expression with decreasing oxygen concentration has also been described for other cell sources (Jiang et al., 1996; Brooks et al., 2009). Inter patient variations in cellular sensitivity to hypoxia are reported (Brooks et al., 2009), which indicate that effects of hypoxia can be more or less pronounced in one patient compared to another. Currently, tissue engineered constructs contain approximately 30% of the amount of collagen associated with native tissue (Balguid et al., 2009). Major improvements in tissue quality and integrity could be obtained if collagen formation and cross linking in tissue engineered constructs can be improved. Native cardiovascular tissue, such as heart valves and blood vessels, mainly consists of collagen type I with smaller amounts of collagen type III. This pattern can also be recognized in the basal expression levels of the myofibroblasts in this study as gene expression levels of procollagen α1(i) were higher than those of procollagen α1(iii). Both levels increased at oxygen levels of 2% or lower. Further lowering of the oxygen concentration had no additional effects. This implies that collagen gene expression increases when oxygen levels drop below a certain threshold. However, below this threshold no additional effects can be obtained by further decreasing the oxygen concentration. In vitro, the effect of hypoxia on collagen expression levels depends on cell source, exposure time and oxygen concentration (Falanga et al., 1993; Agocha et al., 1997; Horino et al., 2002; 63

72 Chapter 4 Hofbauer et al., 2003; Brinckmann et al., 2005). For example, exposure of human dermal fibroblasts to 1% O 2 for 72 hours resulted in a decreased procollagen α1(i) gene expression (Brinckmann et al., 2005), while human dermal fibroblasts and cardiac fibroblasts exposed to 2% O 2 for respectively 96 and 24 hours showed an increase in procollagen α1(i) gene expression (Falanga et al., 1993; Agocha et al., 1997). Furthermore, at oxygen concentrations of 0.5% and 0% O 2 no changes were observed in rat vascular smooth muscle cells (Hofbauer et al., 2003) and fetal rat lung fibroblasts (Horino et al., 2002). We investigated whether the exposure time influenced gene expression, but no differences were observed between 24, 48, 72 and 96 hours of exposure (data not shown). Interestingly, staining for collagen type I revealed no differences, while collagen type III staining was less pronounced at 0.5% O 2. Both are inconsistent with the gene expression results. However, overall collagen synthesis increased in myofibroblasts exposed to 2% and 0.5% O 2 compared to hyperoxia. Both gene expression and immunofluorescent stainings are intracellular measurements at a certain time point, while collagen synthesis measurements represent a cumulative sum of the synthesis during 4 days. Further studies will focus on extracellular collagen deposition for tissue formation and will distinguish between collagen type I and III to investigate if the discrepancy between gene expression and protein level really exists. The enzyme LH2 is essential for collagen cross linking, while LOX is associated with both collagen and elastin cross linking (Kagan, 2000; Bank and van Hinsbergh, 2002). Since strength and maturity of fibers like collagen and elastin are related to the amount of cross links (Balguid et al., 2007), increased amounts of LH2 and LOX could improve tissue integrity. Gene expression levels of the cross link enzymes LOX and LH2 were strongly upregulated by hypoxia reaching an optimum at 0.5% O 2. At protein level, the upregulation of LH2 at 2% and 0.5% O 2 was confirmed. Unfortunately, measurement of LOX at protein level proved unsuccessful. Others showed that upregulation of LOX is also mediated by targets of the hypoxia inducible factors (Wang et al., 2005), which confirm our findings at the gene expression level. Tissue studies will have to be performed to provide more insight on the effect of low oxygen concentrations on crosslink formation, since actual cross link amounts can be measured in tissue engineered constructs. Elastin is currently considered a missing link in tissue engineered valves (Patel et al., 2006; Mol et al., 2009). It is a crucial component in tissue functionality as it is responsible for the resilience of the tissue. Gene expression levels of elastin remained unaffected for all oxygen concentration examined in this study. This might be considered encouraging particularly given the findings of a decreased elastin content when exposing fetal rat lung fibroblasts to hypoxia (Berk et al., 1999; Berk et al., 2005). As soon as techniques to improve elastin production are developed, they might have potential to be combined with the stimulating effect of hypoxia on collagen formation and fiber maturation without affecting newly formed elastin. 64

73 Controlling matrix formation and cross linking by hypoxia Human myofibroblasts remained viable and proliferating at all oxygen concentrations tested in this study. However, cells exposed to 0.5% O 2 for 4 days lost their spindle like shape and proliferation rate decreased, which might indicate cell stress after a prolonged exposure to such a low oxygen concentration. Therefore, 0.5% O 2 might not be suitable for long term tissue culture. At very low oxygen concentrations cells will inevitable convert to anaerobe glycolysis for ATP production, which eventually can lead to acidification of the medium (Brown et al., 2008). In our experiments acidification was not observed, but in long term cultures at 0.5% O 2 this might affect cell viability. Gene expression and protein levels at 2% O 2 increased while cells retained normal morphology and proliferation rate, which makes 2% O 2 a very promising oxygen concentration to improve tissue composition. Finally, gene expression levels of procollagen α1(i) and (III), and LOX were not affected by 4% O 2, but VEGF expression increased, indicating that this oxygen concentration is hypoxic for the cells. Therefore, it might be that culturing at 4% O 2 will lead to increased ECM production in long term experiments. The results of this study show that hypoxia induces ECM formation and maturation by human myofibroblasts. Implementing these results in 3D tissue engineering will help to develop more mature and, therefore, more functionally competent tissue in vitro. However, to implement these results into 3D cardiovascular tissue engineering approaches the oxygen gradient that exists within the cultured tissue must be considered (Brown et al., 2007). Previous 3D tissue culture at 7% O 2 showed a significant increase in the mechanical properties of the tissue. However, due to poor oxygen diffusion, the actual oxygen concentration in the centre of the engineered constructs was probably below 7% O 2. Therefore, oxygen diffusion should be improved to keep this gradient within a 3D construct to a minimum. In addition, numerical simulations might be helpful to assess the environmental oxygen concentration needed to obtain the desired oxygen concentration within the tissue (Sengers et al., 2005). In conclusion, this study highlights the potential improvements in cardiovascular tissue engineering by culturing under hypoxic conditions. Hypoxia enhanced gene expression and protein production of ECM components in human myofibroblasts cultured in 2D when oxygen levels decreased below 4%. Effects were most abundant for the cross link enzymes LOX and LH2, but also collagen synthesis increased at low oxygen concentrations. Implementation of these results in 3D cardiovascular tissue engineering can further improve tissue formation and maturation, which will be valuable in controlling tissue culture and matrix composition towards engineered cardiovascular human implants close to native. 65

74 Chapter Acknowledgements The authors gratefully acknowledge the support of the Smart Mix Program of the Netherlands Ministry of Economic Affairs and the Netherlands Ministry of Education, Culture and Science. The authors would like to thank Marina van Doeselaar for technical assistance and Mariska Huijerjans for performing the pilot experiments. 66

75 Chapter 5 Low oxygen concentrations impair tissue development in tissue engineered cardiovascular constructs This chapter is based on: Marijke A.A. van Vlimmeren, Anita Driessen Mol, Cees W.J. Oomens, Marloes W.J.T. van den Broek, Reinout Stoop, Carlijn V.C. Bouten, Frank P.T. Baaijens (2011) Low oxygen concentrations impair tissue development in tissue engineered cardiovascular constructs. Tissue Eng Part A. (in press)

76 Chapter Introduction The tissue engineering approach is investigated extensively for application in cardiovascular diseases, such as heart valve dysfunction (Flanagan et al., 2009; Mol et al., 2009; Sacks et al., 2009) and coronary artery disease (L'Heureux et al., 2007; Stekelenburg et al., 2009; Tschoeke et al., 2009). Current available implants to replace diseased heart valves and coronary arteries have the disadvantage that they either need anticoagulation treatment to prevent thromboembolism or do not provide a life time solution due to poor durability (Pibarot and Dumesnil, 2009). Cardiovascular tissue engineering seeks to overcome these problems by creating a living and autologous implant, that will be able to grow, adapt and remodel throughout a patient s life, without the need of anticoagulation therapy. Although first in vivo results at pulmonary position have shown promising results (Hoerstrup et al., 2000; Hoerstrup et al., 2006), the collagen network, which is the main load bearing component of native tissues, still needs further improvement to bear the significantly higher loads of the systemic circulation. In particular, collagen cross links are a very important component in the collagen network as they provide strength and stability to the collagen fibers (Balguid et al., 2007). Thus major improvements in tissue integrity will be reached if collagen formation and cross linking in tissue engineered constructs can be improved. One of the environmental factors that might influence collagen formation and cross linking is the oxygen concentration during culture. As the physiological oxygen levels in blood range from 5% O 2 in the venous blood to 13% O 2 in the arterial blood (Guyton and Hall, 2000), normal cell culture at 21% O 2 actually represents a hyperoxic environment for the cells. Previous culture of human vascularderived cells seeded onto a biodegradable scaffold (3D) and cultured at the physiological oxygen concentration of 7% has shown increased tissue strength and cross link density compared to culturing at 21% O 2 (Balguid et al., 2009). To investigate whether culturing at even lower oxygen concentrations could further improve tissue formation, we have recently investigated the effect of culturing at oxygen concentrations of 7% and below on human vascular derived cells (van Vlimmeren et al., 2010). At gene expression level, oxygen levels below 4% O 2 were beneficial for collagen type I and III, and the cross link enzymes lysyl oxidase (LOX) and lysyl hydroxylase 2 (LH2). In addition at protein levels, overall collagen synthesis and LH2 protein amount increased at both 2% and 0.5% O 2 compared to 21% O 2. The aim of the present study was to implement the results obtained at the cellular level into 3D tissue engineering to investigate if culturing at oxygen concentrations below 7% could indeed improve tissue quality. Based on the 2D cell study, we would predict that cells exposed to oxygen concentrations below 4% to increase collagen and collagen cross link production (van Vlimmeren et al., 2010). To estimate cellular oxygen levels in 3D tissues, a finite element model has been developed in which the oxygen gradient within the tissues was determined. For the experimental 68

77 Low oxygen concentrations impair tissue development work, tissue engineered cardiovascular constructs derived from human vascular derived cells seeded onto a rapid degrading scaffold were cultured at 7%, 4%, 2% and 0.5% O 2 and compared to control cultures at 21% O 2 at gene expression and tissue level. These oxygen concentrations were chosen such that the mean oxygen concentrations within the tissue engineered constructs were ~4%, 2%, 0.5% and below 0.5% O 2, respectively. The culture environment was evaluated closely, by measuring and modeling the applied oxygen concentrations and monitoring quality of the culture medium. 5.2 Materials & Methods The theoretical model A theoretical model was used to simulate the oxygen diffusion in tissue engineered (TE) constructs at time of equilibrium of the culture medium with the environment. Since tissues were cultured on a shaking table, it was assumed that the medium is fully homogenized, which allows for modeling only the construct itself. The diffusion of oxygen through the TE construct was described by the diffusion equation (Oomens et al., 2009): c t x c D x q (5.1) where c is the oxygen concentration in the TE construct [nmol/mm 3 ] and D is the diffusion constant of the TE construct [mm 2 /s]. q is the oxygen consumption of the cells [nmol/mm 3 /s], which depends on the amount of oxygen that is present, and can be described by the Michaelis Menten relationship: q V c max k c cell (5.2) where V max is the maximum oxygen consumption rate [nmol/s/cell], K m is the oxygen concentration at which the consumption has dropped to 50% of its original value [nmol/mm 3 ] and ρ cell is the cell density [cells/mm 3 ]. The equations were solved with a custom made finite element package written in Matlab (The MathWorks, Eindhoven, The Netherlands). An axi symmetrical finite element mesh was used to model the TE construct as a circular construct with a contact surface and thickness similar to that of the rectangular TE constructs. The essential boundary conditions at the upper and lower surfaces were set to the oxygen concentrations applied in the experiment (c applied ) in nmol/mm 3. Correcting 21% O 2 in air for 5% CO 2 in the incubator, assuming the air to be fully saturated with water (vapor pressure = atm), and an oxygen solubility of 0.98 m 69

78 Chapter 5 nmol/mm 3 /atm at 37 C (Morsiani et al., 2001), the dissolved O 2 in the medium is equivalent to 183, 64, 37, 18 and 5x10 3 nmol/mm 3 for 21%, 7%, 4%, 2% and 0.5% O 2, respectively. The diffusion constant of the tissue was assumed 2.0x10 9 mm 2 /s based on the diffusion constant of matrigel (Radisic et al., 2006). V max was set to 5x10 8 nmol/s/cell (Casey and Arthur, 2000; Wang et al., 2009) and K M to 6x10 3 nmol/mm 3 (Casey and Arthur, 2000; Radisic et al., 2006; Wang et al., 2009). Simulations were performed with cell density at initial seeding (15 x10 3 cells/mm 3 ) and cell density after four weeks of culture. A sensitivity analysis was conducted on the main model parameters, D, V max, K m and ρ cell, by changing them to times their baseline values. An overview of all constants within the model is provided in table 5.1. Table 5.1: Summary of model parameters Model parameter Definition Value Units Reference c Oxygen concentration within the TE construct nmol/mm 3 D Diffusion coefficient of oxygen in TE construct mm 2 /s Radisic et al q Oxygen consumption rate of the cells nmol/mm 3 V max K m Maximum oxygen consumption rate Oxygen tension at half maximum oxygen consumption rate /s nmol/s/cell Wang et al Casey et al nmol/mm 3 Radisic et al.2006 Wang et al Casey et al ρ cell Cell density (day 0) cells/mm (day 28, 21% O 2 ) cells/mm (day 28, 7% O 2 ) cells/mm (day 28, 4% O 2 ) cells/mm (day 28, 2% O 2 ) cells/mm (day 28, 0.5% O 2 ) cells/mm 3 c applied Applied oxygen concentration (21% O 2 ) nmol/mm (7% O 2 ) nmol/mm (4% O 2 ) nmol/mm (2% O 2 ) nmol/mm (0.5% O 2 ) nmol/mm Cell and tissue culture Vascular derived cells were harvested from the human vena saphena magna obtained according to the Dutch guidelines for secondary use of materials. The cells were seeded onto rectangular shaped scaffolds (18x5x1 mm) of rapid degrading nonwoven polyglycolic acid (PGA) (Concordia Manufacturing Inc, Coventry, RI, USA) coated with poly 4 hydroxybutyrate (P4HB) (Tepha, Lexington, Massachusetts; received as part of the collaboration with University Hospital Zurich). The scaffolds were cultured constrained, a state shown to provide good tissue formation (Mol et al., 2006), by 70

79 Low oxygen concentrations impair tissue development attachment of both ends to a ring of stainless steel with polyurethane tetrahydrofuran glue (15% wt/vol). Sterilization was achieved by 70% ethanol incubation for 30 minutes. Cells were seeded at passage 6 with a seeding density of 15 million cells per cm 3 using fibrin as a cell carrier (Mol et al., 2004). These tissue engineered (TE) constructs were cultured at 21%, 7%, 4%, 2% and 0.5% O 2 for 48 hours (gene expression, n=10) or 4 weeks (tissue formation, n=11 for 21% and n=6 for 7%, 4%, 2% and 0.5%) in a CO2/O2 controllable incubator (MCO 18M SANYO; Wood Dale, USA, 5% CO2, 37 C). Medium was replaced three times a week and consisted of advanced Dulbecco s Modified Eagle Medium (DMEM; Invitrogen, Carlsbad, USA), supplemented with 10% Fetal Bovine Serum (FBS; Greiner Bio one, Frickenhausen, Germany), 1% GlutaMax (Invitrogen), 1% penicillin/streptomycin (Lonza, Basel, Switzerland) and L ascorbic acid 2 phosphate (0.25 mg/ml; Sigma, St. Louis, MO, USA) Gene expression Gene expression levels of vascular endothelial growth factor (VEGF), procollagen α1(i) (col I) and α1(iii) (col III), and cross link enzymes lysyl oxidase (LOX) and lysyl hydroxylase 2 (LH2) were measured in TE constructs. After 48 hours of exposure to each of the five oxygen concentrations, the TE constructs (2 runs of n=5 per oxygen concentration) were washed twice with PBS and stored at 80 C. The samples were mashed in RLT buffer containing 1% β mercaptoethanol (Merck, Schiphol rijk, The Netherlands) with a Mikro Dismembrator S (Sartorius, Nieuwegein, The Netherlands). RNA was isolated using the Qiagen Micro Kit according to the manufacturer s instructions (Qaigen, Hilden, Germany). Synthesis of cdna was carried out with 500 ng RNA using M MLV reverse transciptase (Invitrogen). Gene expression analysis was performed on a icycler Real Time PCR detection system (MyiQ, Bio Rad, Hercules, USA) using iq TM SYBR Green Supermix (Biorad). Cytochrome c 1 (CYC1) was used as reference gene. The primer sequences have been described previously(van Vlimmeren et al., 2010) Evaluation of the tissue culture environment The duration of the temporary increase in oxygen concentration, due to refreshment of medium under atmospheric conditions, was measured in the incubator and culture medium with an InPro 6800 oxygen sensor (Mettler Toledo, Tiel, The Netherlands). Quality control of the culture environment was further performed by measuring the ph and the concentration of lactate and glucose For this, supernatants (n=6 per oxygen concentration) were collected once a week, 3 days after the medium was changed. Measurement of ph was performed by application of ph indicator paper (Merck) ranging ph and ph with unit intervals. Glucose concentration in the medium was measured using an OneTouch UltraEasy blood glucose 71

80 Chapter 5 meter (Life Scan Inc., Milpitas, USA). For lactate production, one part of diluted medium sample was mixed with 10 parts of lactate reagent (Kordia Life Sciences, Leiden, The Netherlands). Absorbance was measured after 10 minutes at 540 nm against background at 650 nm on a plate reader (Synergy HT microplate reader, Winooski, USA, Bio Tek). Calibration curves were obtained for each measurement from a range of dilutions of L (+) lactic acid (Sigma) Mechanical characterization Mechanical properties of the TE constructs (n=5 per oxygen concentration) were assessed with an uniaxial tensile tester (Kammrath & Weiss, Dortmund, Germany) equipped with a 20N load cell. Tensile tests were done at a speed of 100% per minute. Initial width and thickness of the samples were assessed with a SensoFar PLµ 2300 optical imaging profiler (SensoFar Tech, Barcelona, Spain). The Cauchy stress was defined as the force divided by the deformed cross sectional area, assuming incompressible material behavior and plotted against strain. The ultimate tensile stress (UTS) was defined by the peak stress of the obtained stress strain curves. The Young s modulus was determined from the slope of the linear part of the stress strain curve at approximately 15% strain. Surface compaction was defined as the percentage of shrinkage compared to the original cross sectional area Qualitative tissue analyses To analyse tissue quality by histological staining, the TE constructs were fixated with 3.7% formaldehyde (Merck) and embedded in paraffin. Tissue sections of 10 µm thickness were cut and studied by hematoxylin and eosin (HE) stainings for general tissue morphology and Masson Trichrome (MT) staining for deposition of collagen Western blot analysis Lyophilized TE constructs (n=4 per oxygen concentration) were dissolved in RIPA buffer (Sigma) and homogenized by sonification to a normalized concentration of 10 mg dry weight/ml. Proteins (10 µg) were separated by SDS PAGE (Mini PROTEAN 3 System, BioRad, Hercules, USA) conditions on a 10% poly acrylamide gel and transferred to a PVDF membrane (Millipore Corporate, Billerica, MA, USA). The membrane was blocked by incubation in PBS/0.1% Tween 20/5% milk solution for 1 hour and probed overnight at 4 C with either rabbit anti human LH2 (clone SN832, Quickzyme BioSciences, Leiden, the Netherlands) or mouse anti human α smooth muscle actin (α SMA, Sigma) in PBS/0.1% Tween 20/1% milk. After washing and incubation with respectively the conjugates goat anti rabbit peroxidase and rabbit anti mouse peroxidase (Pierce, 72

81 Low oxygen concentrations impair tissue development Rockford, IL, USA) in PBS/0.1% Tween 20 for 1 hour the blots were developed using Supersignal West Dura chemiluminescent Substrate (Pierce) and visualized by Versadoc (Bio Rad). Since reference proteins were affected by culturing at low oxygen concentration, the percentage of protein loaded onto the gel from the total protein present in the TE construct has been added in the results to show that the differences are apparent Quantitative tissue analyses To quantify tissue formation after 4 weeks, the amount of DNA, sulfated glycosaminoglycans (GAGs), hydroxyproline; and lysylpyridinoline (LP) and hydroxylysylpyridinoline (HP) cross links was measured (n=5 per oxygen concentration). Lyophilized tissue samples were digested in papain solution (100 mm phosphate buffer, 5 mm L cysteine, 5mM ethylenediaminetetraacetic acid (EDTA), and µg papain per ml) at 60 C for 16 hours. DNA content was determined with the Hoechst dye method (Cesarone et al., 1979) and a standard curve of calf thymus DNA (Sigma). The total amount of cells within the construct was calculated based on the assumption that every cell contains 6.5 pg of DNA (Dolezel et al., 2003). GAG content was determined with a modification of the assay described by Farndale et al. (Farndale et al., 1986) and a standard curve from chondroitin sulfate from shark cartilage (Sigma). In short, 40 μl of diluted sample, without addition of chondroitin AC lyase, chondroitin ABC lyase and keratanase, was pipetted into a 96 well plate in duplicate. Subsequently, 150 μl dimethylmethylene blue was added and absorbance was measured at 540 nm. For collagen analyses, the digested samples were hydrolyzed in 6M hydrochloric acid (Merck). Hydroxyproline residues were measured on the acid hydrolysates using reverse phase high performance liquid chromatography after derivatization with 9 fluorenylmethyl chloroformate (Bank et al., 1996). Hydroxyproline content was converted to collagen by assuming that a triple helix consists of 300 hydroxyproline residues. The same hydrolysates were used to measure the number of LP and HP crosslinks, using high performance liquid chromatography as described previously (Robins et al., 1996; Bank et al., 1997) Statistical analyses All data are presented as mean and their standard error of mean. Relative gene expression levels were normalized to the reference gene and the control group (21% O 2 )(Hellemans et al., 2007; Bustin et al., 2009). One way ANOVA, followed by a Tukey s multiple comparison post hoc test, was carried out to compare oxygen concentrations to control cultures at 21% O 2. Differences were analyzed using GraphPad Prism software (GraphPad Software, Inc, USA) and considered significant for P values <

82 Chapter Results The theoretical model and the calculated oxygen concentrations in tissue engineered constructs Figure 5.1 and table 5.2 show the results of the theoretical model and highlight considerable differences between the oxygen concentration applied to the environment and the oxygen concentration to which the cells are exposed. The desired oxygen concentrations of 4%, 2%, 0.5% and below 0.5% O 2, led to tissue culture at 7%, 4%, 2% and 0.5% O 2, respectively (Table 5.2). TE constructs cultured at 21% O 2 served as control. Figure 5.1: The oxygen concentration at day 0 (solid line) and day 28 (dashed line) throughout the thickness of TE constructs cultured at 21%, 7%, 4%, 2% and 0.5% O 2. The oxygen concentrations were lower at day 28 than at day 0 for all oxygen concentrations except 0.5% O 2. 74

83 Low oxygen concentrations impair tissue development During culture, the average oxygen concentrations changed due to changes in cell number. In figure 5.1, the dotted line represents the oxygen concentration after 4 weeks of culture. For tissue culture at 21%, 7%, 4% and 2% O 2, the oxygen concentration within the TE construct decreased. This decrease became less with decreasing oxygen concentration, until, at 0.5% O 2, a small increase was observed due to decreased cell number. Table 5.2: Minimum and averaged oxygen concentrations in TE constructs cultured at different oxygen concentrations. A sensitivity analysis of the theoretical model (Figure 5.2) indicated that the model is very sensitive to an increase in the cellular consumption (V max ) and amount of cells (ρ cell ), and to a decrease in the diffusion coefficient (D). By contrast, the model is less sensitive to changes in the oxygen tension at half maximum oxygen consumption rate (K m ). Figure 5.2: The influence of increasing and decreasing the model parameters D, V max, K m and ρ cell ten times on the minimum oxygen concentration in TE constructs cultured at 21% O 2. The model is very sensitive to changes in D, V max and ρ cell and less sensitive to changes in K m Gene expression An overview of the normalized gene expression levels is provided in figure 5.3. VEGF gene expression levels increased at 4% and 2% compared to 21% O 2. LH2 and LOX levels were elevated at 7%, 4% and 2% O 2, but only gene expression of LOX at 7% O 2 was significantly higher than 21% O 2. Gene expression levels of procollagen α1(i) and procollagen α1(iii) remained unaffected by all oxygen concentrations. 75

84 Chapter 5 Figure 5.3: Normalized gene expression of 5 genes from TE constructs exposed to 21%, 7%, 4%, 2% and 0.5% O 2. VEGF gene expression levels increased at 4% and 2% O 2 compared to 21% O 2. Gene expression of LOX was increased at 7% compared to 21% O Culture environment The applied oxygen concentration was reached within 30 minutes inside the incubator and after hours in the culture medium. Figure 5.4 gives an overview of the ph, glucose consumption and lactate production in medium supernatants exposed to TE constructs for 3 days. During the four weeks of culture, ph at 21%, 7%, 4% and 2% O 2 decreased and glucose consumption and lactate production increased. At 0.5% O 2, changes over time were only sparsely detected, while ph was higher and glucose consumption and lactate production were lower than 21% O 2 at all time points investigated. Compared to 21% O 2, increased glucose consumption and lactate production was observed at 7%, 4% and 2% O 2 in week 1 and 2, and decreased ph was observed in week 2 at 4% O 2. Glucose consumption increased and ph decreased compared to 21% at 2% in week 3 and at 4% in week Qualitative tissue evaluation Macroscopic images and histological staining of sections of the TE constructs are shown in figure 5.5. After 4 weeks of culture, TE constructs at 21%, 7% and 4% O 2 showed homogeneous tissue formation with comparable tissue density and abundant collagen formation (Figure 5.5A C, F H, K M). By contrast, TE constructs cultured at 2% O 2 were less dense with less collagen present (Figure 5.5D,I,N), while TE constructs cultured at 0.5% O 2 showed hardly any tissue formation and collagen (Figure 5.5E,J,O). 76

85 Low oxygen concentrations impair tissue development Figure 5.4: ph (A), glucose consumption (B) and lactate production (C) in supernatants from TE constructs at week 1, 2, 3 and 4 (n=6 per concentration per week). At 0.5% O 2, ph was higher and glucose consumption and lactate production were lower than 21% O 2. In week 1 and 2, glucose consumption and lactate production increased at 7%, 4% and 2% compared to 21% O 2. ph at 4% O 2 decreased in week 2. In week 3, glucose consumption increased and ph decreased at 2% O 2. In week 4, glucose consumption increased and ph decreased at 4% O Mechanical properties Figure 5.6A represents the averaged stress strain curves of the TE constructs and demonstrates non linear tissue behavior for constructs cultured at 21%, 7%, 4% and 2% O 2. This clearly reflects a toe in region at low stress level, followed by the linear region up to the UTS value. Tissue stiffness (E modulus), strength (UTS) and surface compaction at 7% and 4% O 2 were similar to those at 21% O 2 (Figure 5.6B D). At 2% O 2, stiffness decreased compared to TE constructs at 21% O 2, although strength and surface compaction remained similar (Figure 5.6B D). At 0.5% O 2, the limited tissue formation resulted in undetectable force measurements and less surface compaction (Figure 5.6B D). 77

86 Chapter 5 Figure 5.5: Macroscopic pictures (A E), hematoxylin & eosin staining (F J) and Masson Trichrome staining (K O) of TE constructs cultured for 4 weeks at 21% (A,F,K), 7% (B,G,L), 4% (C,H,M), 2% (D,I,N) and 0.5% (E,J,O) O 2. Collagen is shown in blue, cytoplasm in pink/red and nuclei in purple/black. TE constructs cultured at 21%, 7% and 4% O 2 showed dense tissue formation with similar amounts of collagen. At 2% O 2, tissue was less dense and collagen amount was little. At 0.5% O 2, hardly any tissue and collagen was formed. The black bar represents 200 µm. 78

87 Low oxygen concentrations impair tissue development Figure 5.6: Stress strain curves (A), E modulus (B) UTS (C) and surface compaction (D) of TE constructs cultured at 21%, 7%, 4%, 2% and 0.5% O 2 for 4 weeks. Mechanical properties at 7% and 4% O 2 were similar to 21% O 2. At 2% O 2, the E modulus decreased compared to 21% O 2 and at 0.5% O 2, both UTS and E modulus decreased compared to 21% O 2. Surface compaction at 0.5% was less compared to 21% O 2. * and ** denote a significant difference of respectively p<0.05 and p<0.001 compared to controls cultured at 21% O Tissue composition Figure 5.7 and table 5.3 give an overview of the tissue composition at the different oxygen concentrations and significant differences compared to 21% O 2. DNA content and the associated cell number per TE construct indicated that cell proliferation occurred in the tissues culture at 21%, 7%, 4% and 2% O 2, while cell death was apparent at 0.5% O 2. Decreased amounts of GAGs per TE construct were observed at 7%, 2% and 0.5% O 2 compared to 21% O 2, but GAGs normalized to DNA did not change (Table 5.3). The amount of collagen per strip decreased at 2% and 0.5% O 2 compared to 21% O 2. Collagen production normalized to DNA did not decrease at 2% O 2, but was very little at 0.5% O 2 (Table 5.3). The amount of mature HP cross links was similar at 21%, 7%, 4% and 2% O 2, while the amount of LP cross links increased at 4% and 2% compared to 21% O 2. Both HP and LP cross links were undetectable at 0.5% O 2. 79

88 Chapter 5 Figure 5.7: DNA (A), GAG (B), collagen (C), and HP (D) and LP (E) cross links per triple helix content per TE construct cultured at 21%, 7%, 4%, 2% and 0.5% O 2. The black line in (A) indicates the amount of cells seeded at day 0. At 7% O 2, GAG content decreased compared to 21% O 2. At 4% O 2, increased amounts of LP cross links were observed. At 2% O 2, GAG and collagen content decreased and the amount of LP cross links increased. At 0.5% O 2, cellular, GAG and collagen content decreased. * and ** denote a significant difference of respectively p<0.05 and p<0.001 compared to controls cultured at 21% O 2. Table 5.3: Tissue composition and mechanical properties in TE constructs cultured at different oxygen concentrations. Table 5.3: * and ** denote a significant difference of respectively p<0.05 and p<0.001 compared to controls cultured at 21% O 2. 80

89 Low oxygen concentrations impair tissue development Western blot (Figure 5.8) revealed that the cross link enzyme LH2 was present in all TE constructs, but were most pronounced at 21% O 2. Cells in all TE constructs expressed similar amounts of α SMA for all oxygen concentrations, except for 0.5% O 2, for which little α SMA expression was present. Figure 5.8: (A) LH2 was present in all TE constructs, but most pronounced at 21% O 2. α SMA was present in all constructs, except for 0.5% O 2. (B) The percentage of protein from the total amount of protein present in the TE engineered construct loaded onto the gel. At 0.5% O 2, more protein was loaded onto the gel, indicating the absence of α SMA at this oxygen concentration to be apparent. 5.4 Discussion In vivo functionality of tissue engineered cardiovascular implants, such as heart valves and blood vessels, requires the in vitro development of a well organized and functional tissue. In the present study, a range of oxygen concentrations was tested for their potential to improve tissue integrity. A previously performed study within our group suggested an increase in collagen cross link content and strength when culturing at 7% O 2 (Balguid et al., 2009). To investigate if lower oxygen concentrations could further improve tissue formation, we have cultured vascular derived cells at a range of oxygen concentrations below 7% (van Vlimmeren et al., 2010). This demonstrated that cross link enzymes exponentially increase with decreasing oxygen concentrations down to 0.5% O 2, while collagen increase remained constant once the oxygen concentration decreased below 2% O 2. In the present study TE constructs have been cultured at 7%, 4%, 2% and 0.5% O 2. Culturing at 7% O 2 results in a mean oxygen concentration of 4.4% O 2, which 81

90 Chapter 5 according to the 2D cell study induces minimal increase in collagen cross link formation. Culturing at 4% O 2 was predicted to improve the previous results at 7%, as the average oxygen concentration drops to 2.0% which should increase both collagen and collagen cross link amounts. At 2% O 2, the average oxygen concentration becomes 0.8% O 2, which was predicted to further increase collagen cross link formation. Finally, 0.5% O 2 was predicted to be too severe for long term tissue culture. Gene expression results showed increased expression levels of VEGF at 4% and 2% O 2, indicating cellular hypoxia at those levels. At 0.5% O 2 VEGF expression was very low, which might be due to the severity of that condition. Gene expression levels of the collagen cross link enzymes LOX and LH2 increased at 7%, 4% and 2% O 2, but there was some variability and thus only the increase of LOX at 7% O 2 was significant. These results emphasize an issue of reproducibility within these cell based of experiments, but also indicate that hypoxia might have potential to improve cross link formation. However, these short term increases of gene expression did not translate to protein production in the long term tissue culture experiments. TE constructs cultured at 21%, 7% and 4% O 2 showed dense and homogeneous tissue formation with similar values for strength, stiffness and collagen and collagen cross link content. By contrast, at 2% O 2, collagen content, GAG content, and stiffness all decreased. At 0.5% O 2 very little tissue was formed. Overall, tissue properties deteriorated at the lowest oxygen concentrations, which was in contrast with the present hypothesis based on the results of the 2D monolayer and 3D tissue study of Balguid et al., and the gene expression results in the present study. It appears that there is a discrepancy between short term cell culture and long term tissue culture experiments that causes differences in cell behavior. Increased matrix metalloproteinase (MMP) activity could reduce the final collagen content, even if collagen synthesis levels remain constant or increase. In human cardiac myofibroblasts MMP activity reduced at hypoxia (Morley et al., 2007; Riches et al., 2009), which suggests that MMP activity would rather decrease than increase. However, as differences between cell sources are apparent, the effect of hypoxia on MMP activity should be measured for vascularderived cells to investigate if this explains the discrepancy between short term and longterm culture. Anaerobic cell metabolism could influence long term tissue culture at low oxygen concentrations, as it results in substantial glucose consumption and lactic acid production which causes acidification of the cellular environment (Alberts et al., 2002). In the first two weeks lactate production and glucose consumption were higher at 7%, 4% and 2% compared to 21% O 2, indicating increased anaerobic cell metabolism at these lower oxygen concentrations. In week 3 and 4, this difference was less pronounced. Although cells switched to anaerobic metabolism, ph of the culture medium did not decrease below the threshold for ph associated with viable cells, which ranges from

91 Low oxygen concentrations impair tissue development to 7.0 (Rotin et al., 1986; Gawlitta et al., 2007; Brown et al., 2008). However, local acidification in the middle of the constructs, due to poor diffusion of lactic acid away from the tissue, might have impaired tissue formation. Assumptions applied to the theoretical model might have overestimated the oxygen concentrations within the TE constructs. It assumes a homogeneous cell distribution while, in reality, there is a higher cell density at the surfaces of the TE constructs. The cells in this denser surface layer will experience the environmental oxygen concentration while the surface layer may impair oxygen diffusion. In addition, the diffusion constant will change during culture when tissue becomes denser. The diffusion constant is a very sensitive model parameter and oxygen concentrations in the middle of the construct diminish if it decreases. Therefore, obtaining the appropriate diffusion constant matched to the culture time point and implementing inhomogeneous cell distribution will be focus of further development of the model. At 0.5% O 2, significant cell death occurred and the cells that survived were minimally active (little glucose consumption and lactate production) and expressed little α SMA, indicating a change in phenotype. Vascular derived cells can shift from a quiescent, contractile phenotype to a synthetic phenotype, which is characterized by proliferation and ECM synthesis (Rensen et al., 2007; Beamish et al., 2010). The contractile phenotype is associated with α SMA whereas the synthetic phenotype expresses little or no α SMA. Here, lack of α SMA correlated with poor tissue formation, but this was probably caused by cellular stress due to the severe hypoxia. In a previous study, Modaressi et al. observed less contractile properties in dermal myofibroblasts exposed to low oxygen concentrations, resulting in decreased α SMA expression (Modarressi et al., 2010). The significant lower compaction and α SMA expression at 0.5% O 2 observed in the present study correlate with their findings, although the present study did not suggest an increase in tissue formation and maturation. Based on our previously performed experiments on 2D monolayers and 3D tissues (Balguid et al., 2009; van Vlimmeren et al., 2010), an increase in collagen crosslinks at all oxygen concentrations tested was predicted. In the present study, the amount of mature HP cross links was similar at 21%, 7%, 4% and 2% O 2 and undetectable at 0.5% O 2. The amount of LP cross links increased at 4% and 2% O 2. The cross link enzyme LH2, needed to form HP and LP cross links, was present in all TE constructs, but most pronounced at 21% O 2, in contrast to the results at the cellular level (van Vlimmeren et al., 2010). Increased LP cross links could indicate susceptibility to calcification, since LP cross links are primarily found in calcified tissue (Bailey et al., 1998). However, the amount of LP cross links was very low compared to the HP crosslinks. In previous tissue culture at 7% O 2, the amount of HP cross links per triple helix was suggested to double compared to control cultures at 21% O 2 (Balguid et al., 2009).However, new insights on inter experimental variation have indicated that the 83

92 Chapter 5 separate culture of the control group might have affected the conclusions in that study. Eliminating the control group in that study, the effects could have been caused by the combination of insulin and hypoxia. We have recently repeated the experiments of that study to investigate this hypothesis, there were no significant differences between control, hypoxia, insulin or combinations of both in the amount of collagen and collagen cross links, and the mechanical characteristics. In retrospect, we therefore feel that the positive effect of hypoxia that was suggested in the work of Balguid et al. is inaccurate. However, as gene expression results at both 2D and 3D level, do show the potential of hypoxia, we believe that further investigation is definitely important and should focus on intermittent hypoxia. In the present study and previous study (Balguid et al., 2009), intermittent hypoxia was applied to the TE constructs due to medium changes at 21% O 2. In the present study, medium was replaced three times a week, which resulted in a period of hours of increased oxygen concentration per medium refreshment. In the previous study at 7% O 2, medium was changed twice a week. It is not known whether this intermittent period positively triggers cells every time they reach a low oxygen concentration or whether it is too short to induce changes. For human dermal fibroblasts, exposure time appears to determine the response of the cells. Procollagenα1(I) gene expression levels of human dermal fibroblasts exposed to 1% O2 increased with exposure time up to 72 h. After 96 h the increase became less and after chronic exposure (up to 6 passages) it dropped from 8.7 to 1.8 fold (Siddiqui et al., 1996). Therefore, differences in intermittent time frames could have caused the inconsistency between the two studies and should be part of future research. In conclusion, the TE constructs cultured in this study did not benefit from culture at oxygen concentrations below 7%. Indeed tissue properties at 7% and 4% O 2 were similar to 21% O 2, while culturing at 2% and 0.5% O 2 reduced tissue formation and according mechanical properties. This suggests a delicate balance between exposure time and oxygen concentration and hence we do not rule out that culturing at low oxygen concentrations can have positive effects. Further research will focus on unraveling this balance to optimize tissue formation and maturation with the help of an improved theoretical model to accurately assess oxygen concentrations in the tissues in a time dependent manner. 5.5 Acknowledgements The authors gratefully acknowledge the support of the Smart Mix Program of the Netherlands Ministry of Economic Affairs and the Netherlands Ministry of Education, Culture and Science. The authors like to thank Leonie Grootzwagers and Jessica Snabel for performing the biochemical assays. 84

93 Chapter 6 The potential of prolonged tissue culture to reduce stress generation and retraction in engineered heart valve tissues This chapter is based on: Marijke A.A. van Vlimmeren, Anita Driessen Mol, Cees W.J. Oomens, Frank P.T. Baaijens (2011) The potential of prolonged tissue culture to reduce stress generation and retraction in engineered heart valve tissues Tissue Eng Regen Med. (submitted)

94 Chapter Introduction Tissue engineering approaches, in which cells are seeded into a carrier material, are regularly associated with cell traction mediated tissue shrinkage (Chun et al., 2003; Shi and Vesely, 2003; Stegemann and Nerem, 2003; Balestrini and Billiar, 2009; Flanagan et al., 2009). Cells will always exert traction forces on their surroundings to achieve a basal internal tension level (Eastwood et al., 1996; Brown et al., 1998). During tissue culture, they remodel their surroundings until they have reached this internal stress level, resulting in compaction in the non constrained direction and stress generation in the constrained direction (Legant et al., 2009; John et al., 2010). Upon release of constraints, the developed stress and exerted cell traction forces cause retraction of the tissue (Balestrini and Billiar, 2009; Van Vlimmeren et al., 2011a). There is considerable focus on the tissue engineering of heart valves as an alternative for current available heart valve replacements. These valves are manufactured by seeding vascular derived cells onto a rapidly degrading scaffold (Mol et al., 2006). Tissue compaction and retraction are limiting functionality of such valves. Compaction causes leaflet flattening which reduces the coaptation area of the leaflets. Additionally, due to retraction at the time of implantation, the heart valves are not able to close during diastole and this affects their functionality. In vivo, decreasing leaflet length due to tissue retraction has been observed to cause valvular regurgitation in sheep (Flanagan et al., 2009; Gottlieb et al., 2010). We have developed an in vitro model system of rectangular tissue engineered constructs to investigate the development of compaction, stress generation and retraction (Van Vlimmeren et al., 2011a). It was observed that compaction and stress generation develop as soon as the scaffold loses its mechanical integrity. After 4 weeks of culture, compaction was 50%, the internal stress was 8 kpa and retraction was 36% (Van Vlimmeren et al., 2011a). In the case of a tissue engineered heart valve, this degree of retraction would make it impossible to ensure full closing of the valve leaflets. Studies investigating traction forces, compaction and retraction generally focus on short time changes studied in gel systems that do not change composition during the time frame of the study (Eastwood et al., 1996; Brown et al., 1998; Chun et al., 2003; Shi and Vesely, 2003; Stegemann and Nerem, 2003; Wakatsuki and Elson, 2003; Balestrini and Billiar, 2009; Flanagan et al., 2009; Legant et al., 2009; John et al., 2010). However, for tissue engineered heart valves, scaffolds are used that degrade over time, while cells develop their own extracellular matrix, constantly changing the composition of the TE heart valve. The total stress within a tissue is the result of a balance between the traction forces exerted by the cells and the ability of the scaffold and tissue properties to resist these traction forces. Initially, the scaffold material is stiff and able to withstand the traction forces of the cells. However, after week 2, the scaffold rapidly degrades (Klouda et al., 2008), while the newly formed tissue is still not able to resist the traction 86

95 Prolonged tissue culture reduces stress generation and retraction forces of the cells. This results in compaction during culture and retraction at release of constraints (Van Vlimmeren et al., 2011a). We believe that a well developed tissue is able to resist the traction forces of the cells, resulting in less compaction and retraction. Prolonged tissue culture is hypothesized to improve the extracellular matrix properties and therefore improve its resistance to the traction forces of the cells. In our previous study, we cultured for 4 weeks, after which compaction had reached a constant level, but stress had not attained an equilibrium value (Van Vlimmeren et al., 2011a). The goal of the present study was to quantify force and stress generation, compaction and retraction during a prolonged culture period of 8 weeks and correlate this to the tissue properties, in order to establish a balance between tissue properties and traction forces of the cells. A representative model system of tissue engineered strips was used to investigate these parameters during culture and after release of constraints at week 4, 6 and Material & Methods Tissue culture Vascular derived cells were harvested from the human vena saphena magna obtained according to the Dutch guidelines for secondary used materials. Cells were expanded using standard cell culture methods in a humidified atmosphere containing 5% CO 2 at 37 C. Culture medium consisted of advanced Dulbecco s Modified Eagle Medium (DMEM; Invitrogen, Carlsbad, USA), supplemented with 10% Fetal Bovine Serum (FBS; Greiner Bio one, Frickenhausen, The Netherlands), 1% GlutaMax (Invitrogen), and 1% penicillin/streptomycin (Lonza, Basel, Switzerland). The cells were seeded onto rectangular shaped scaffolds (18x5x1 mm) of rapid degrading nonwoven polyglycolic acid (PGA) (Concordia Manufacturing Inc, Coventry, RI, USA) coated with poly 4 hydroxybutyrate (P4HB) (received as part of the collaboration with University Hospital Zurich) as described previously (Mol et al., 2005). In summary, the scaffolds were attached to two sliding blocks with polyurethane tetrahydrofuran glue (15% wt/vol). Sterilization was achieved by 70% ethanol incubation for 30 minutes. Cells were seeded at passage 7 with a seeding density of 15 million cells per cm 3 using fibrin as a cell carrier (Mol et al., 2004). The tissue engineered (TE) constructs were cultured for 4, 6 and 8 weeks in constrained (n=5 per culture period) and for 8 weeks in semiconstrained configuration (n=5) as described below. During tissue culture, the medium was supplemented with L ascorbic acid 2 phosphate (0.25 mg/ml; Sigma, St. Louis, MO, USA) and replaced twice a week. 87

96 Chapter The model system A previously described model system (Van Vlimmeren et al., 2011a) (Figure 6.1), in which TE constructs were cultured in between two sliding blocks, was used to quantify stress generation and retraction from a single TE construct. One sliding block can be either fixed (during culture) or move freely (to measure retraction after culture) (Figure 6.1C). The other is connected to two leaf springs and the displacement of this sliding block is proportional to the generated force within the tissue (Figure 6.1D). To measure force generation during culture, the sliding block of the leaf springs is not fixed. Because the force that is generated by the cells will deform the leaf springs this configuration is referred to as semi constrained. To measure generated force after release of constraints, both sides are fixed and this configuration is referred to as constrained. Each set up was calibrated with an individual force displacement curve. Figure 6.1: Photograph (A) and schematic picture (B) of a model system in which retraction R L (C) and generated force F (D) can be measured from one single TE construct through the displacement of two sliding blocks positioned opposite from each other Quantification of surface compaction, generated stress and retraction The compacted width (W comp ) of the TE constructs was assessed from photographs and compaction (C W ) was defined as the percentage of shrinkage compared to the original width (W 0 ): 88

97 Prolonged tissue culture reduces stress generation and retraction Wcomp 1 100% C (6.1) W W 0 Thickness was measured with a caliper when samples were sacrificed. Time points in between the sacrificing time points were interpolated. From the measured tissue width and interpolated thickness, the change in cross sectional surface area ( ) over time was assessed. Force generation ( F ) and the retracted length of the TE constructs ( L retract ) were quantified in Matlab (The MathWorks, Eindhoven, The Netherlands), based on the displacement of the sliding blocks (Van Vlimmeren et al., 2011a). The compacted crosssectional surface area was used to determine the generated stress ( ) within the tissue using: F (6.2) A Retraction in length R L (%) was defined as the percentage of shrinkage compared to the original length L 0 (µm): Lretract 1 100% R (6.3) L L Experimental design TE constructs were cultured for 8 weeks in constrained (n=15) and semiconstrained (n=5) configuration. During culture, the compacted width of both constrained and semi constrained TE constructs and generated stress in the semiconstrained TE constructs were measured twice a week. After 4, 6 and 8 weeks of culturing, generated stress and retraction were measured in the constrained samples after release of the constraints (n=5 per time point). Generated stress was measured at 3 minute intervals during a total period of 30 minutes. Retraction in length was measured every 6 minutes in the first half hour, followed by measurements after 1, 2, 4, 6, 16 and 24 hours. An overview of the experimental design is given in figure 6.2. The experiment was performed twice, according to the same protocol with similar cells, cell passage, cell seeding density and scaffold batch. Results from the first experiment are presented in graphical form. And from the second experiment, results are presented in a table. To investigate changes over time and differences between groups, both runs have been combined by normalization. 89

98 Chapter 6 Figure 6.2: Schematic overview of the experimental design. Compaction during culture was measured in both constrained and semi constrained samples. Retraction and generated stress were measured at week 4, 6 and 8 after release of constraints in the constrained samples. Generated stress during culture was measured in the semiconstrained samples. Retraction after release of constraints was measured in the semiconstrained samples at week 8. Finally, for each group, one TE construct was used for histological staining, while the other four were used for biochemical assays and Western blot Qualitative tissue analyses Tissue formation was evaluated qualitatively by histological staining (n=1 per time point). The TE constructs were fixated with 3.7% formaldehyde (Merck) and embedded in paraffin. Tissue sections of 10 µm were cut and studied by hematoxylin and eosin (HE) staining for general tissue morphology and Masson Trichrome (MT) staining for deposition of collagen and Alcian blue staining for deposition of GAGs. The stained sections were visualised using light microscopy. (Axio Observer, Zeiss, Göttingen, Germany) Western blot analysis Lyophilized TE constructs (n=4 per group) were dissolved in RIPA buffer (Sigma) and homogenized by sonification to a normalized concentration of 10 mg dry weight/ml. Proteins (10 µg) were separated by SDS PAGE (Mini PROTEAN 3 System, BioRad, 90

99 Prolonged tissue culture reduces stress generation and retraction Hercules, USA) conditions on a 10% poly acrylamide gel and transferred to a PVDF membrane (Millipore Corporate, Billerica, MA, USA). The membrane was blocked by incubation in PBS/0.1% Tween 20/5% milk solution for 1 hour and probed overnight at 4 C with either mouse anti human α smooth muscle actin (α SMA, Sigma) or mouse anti human β actin (Abcam, Cambridge, UK) in PBS/0.1% Tween 20/1% milk. After washing and incubation with the conjugate rabbit anti mouse peroxidase (Pierce, Rockford, IL, USA) in PBS/0.1% Tween 20 for 1 hour the blots were developed using Supersignal West Dura chemiluminescent Substrate (Pierce) and visualized by Versadoc (Bio Rad) Quantitative tissue analyses To quantify tissue formation, the amount of DNA, sulfated glycosaminoglycans (GAGs), collagen; and hydroxylysylpyridinoline (HP) cross links was measured (n=4 per group). Lyophilized tissue samples were digested in papain solution (100 mm phosphate buffer, 5 mm L cysteine, 5mM ethylenediaminetetraacetic acid (EDTA), and µg papain per ml) at 60 C for 16 hours. DNA content was determined with the Hoechst dye method (Cesarone et al., 1979) and a standard curve of calf thymus DNA (Sigma). The total amount of cells within the construct was calculated based on the assumption that every cell contains 6.5 pg of DNA (Dolezel et al., 2003). GAG content was determined with a modification of the assay described by Farndale et al. (Farndale et al., 1986) and a standard curve from chondroitin sulfate from shark cartilage (Sigma). ). In short, 40 μl of diluted sample, without addition of chondroitin AC lyase, chondroitin ABC lyase and keratanase, was pipetted into a 96 well plate in duplicate. Subsequently, 150 μl dimethylmethylene blue was added and absorbance was measured at 540 nm. For collagen analyses, the digested samples were hydrolyzed in 6M hydrochloric acid (Merck). Hydroxyproline residues were measured on the acid hydrolysates using reverse phase high performance liquid chromatography after derivatization with 9 fluorenylmethyl chloroformate (Bank et al., 1996). Hydroxyproline content was converted to collagen by assuming that a triple helix consists of 300 hydroxyproline residues. The same hydrolysates were used to measure the number of HP cross links, using high performance liquid chromatography as described previously (Robins et al., 1996; Bank et al., 1997) Statistical analyses All data are presented as mean and their standard error of the mean. Measurements over time have only been presented for one run. Both runs have been combined by normalization to the constraint culture group at week 4 to perform statistical analyses. One way ANOVA, followed by a Tukey s multiple comparison posthoc test, was carried out to compare weeks 4, 6 and 8 in the constrained and semi 91

100 Chapter 6 constrained configurations. Analysis were also performed to evaluate correlations between tissue properties, stress generation and retraction. Statistical analyses were done using GraphPad Prism software (GraphPad Software, Inc, USA) and considered significant for P values < Results Tissue formation in prolonged tissue culture Macroscopic photographs of the TE constructs within the model system are shown in figure 6.3A D. Figure 6.3D demonstrates the movement of the leaf springs during culture in the semi constrained configuration. At week 4, thin and homogeneous tissues were formed (Figure 6.3E). At week 6 and 8, a thin fibrous layer was formed around the denser inner tissue which yielded an indistinct outline (white arrows in figure 3F H). Figure 6.3: Macroscopic pictures of the model system (A D) and TE constructs (E H), hematoxylin & eosin staining (I L), Masson Trichrome staining (M N) and Alcian Blue staining (Q T) after 4, 6 and 8 weeks of culture in constrained and semi constrained configuration. Collagen is shown in dark blue, cytoplasm in pink/red, nuclei in purple/black and GAGs in bright blue. At week 4 thin and dense tissues were formed with abundant collagen and GAG formation. At week 6 and 8, a fibrous surface layer was formed around the tissue (indicated by the white arrows). The white and black bars represent respectively 5mm and 200 µm. 92

101 Prolonged tissue culture reduces stress generation and retraction The HE stainings (Figure 6.3I L) show dense and thin tissue at week 4, which becomes thicker at week 6 and less dense at week 8. The MT (Figure 6.3M P) and Alcian blue (Figure 6.3Q T) stainings indicate abundant collagen and GAG formation at all time points. Both stainings show the surface layer that was formed at week 6 and 8, which is indicated by white arrows. No differences between the constrained and semiconstrained configurations were observed at week 8. Cells expressed similar amounts of α SMA in both configurations and all time points (Figure 6.4). Figure 6.4: The amount of α SMA expressed by the cells remained constant over time and was similar between constrained and semi constrained configuration. The amounts of DNA, GAG, collagen and collagen cross links of both runs are presented in table 6.1. Normalized quantitative content in the TE construct is shown in figure 6.5. The amount of DNA, collagen and collagen cross links in the constrained samples was constant over time. By contrast, the amount of GAGs increased at week 6 and 8 compared to week 4 (Figure 6.5D). No differences in tissue properties were observed between the constrained and semi constrained configurations at week Tissue compaction and generated stress during culture (semiconstrained configuration) Tissue compaction was evident at week 2 and continued up to week 4, at which point its value was 63.9±0.8% (Figure 6.6A). Beyond this time, its value decreased to 59.1±1.8% and 55.0±2.0% at week 6 and 8, respectively. The width of the specimen could be measured easily by the use of a camera system. Measuring the thickness, however, was more difficult and was only performed at week 4, 6 and 8 resulting in averaged thicknesses of 700±40, 2000±20 and 1900±100 µm, respectively. To be able to see a trend in the surface change, we interpolated the time points, resulting in the 93

102 Chapter 6 temporal profile indicated in figure 6.6B. The findings indicate that tissue thickness decreased during the first 4 weeks, then rapidly increased up to week 6 and remained stable beyond that time. Figure 6.5: DNA (A), collagen (B), collagen cross link (C), and GAG (D) content per TE construct cultured for 4, 6 and 8 weeks in constrained and 8 weeks in semi constrained configuration. The amount of DNA, collagen and collagen cross links remained constant over time. The amount of GAGs per TE construct increased over time. No differences were observed in tissue composition between constrained and semi constrained tissue culture at week 8. * and ** denote a significant difference of respectively p<0.05 and p<0.001 compared to constrained culture at week 4. Combining thickness with the measured changes in width over time (Figure 6.6B), a time profile of the cross sectional area was estimated (Figure 6.6C). The crosssectional area drastically decreased from 5.85±0.18 mm 2 to 1.58±0.10 mm 2 in the first 4 weeks, dominated by changes in width, after which it increased to 4.05±0.38 mm 2 at week 6, dominated by changes in thickness. Finally, at week 8, the mean cross sectional surface was 5. 13±0.46 mm 2. The force generated by the cells in the semi constrained samples increased to a value of 44±4 mn at week 4, after which it remained constant with averaged values of 43±5 and 41±5 mn at weeks 6 and 8, respectively (Figure 6.6D) The generated stress, however, was found to increase up to 27.8±3.2 kpa at week 4, after which it rapidly decreased to 11.7±1.3 kpa at week 6 and decreased further to 8.9±1.1 kpa at week 8 (Figure 6.6E) 94

103 Prolonged tissue culture reduces stress generation and retraction Figure 6.6: Tissue compaction (A C), force (D) and stress (E) generation during culture. (A) TE constructs started to compact at week 2, continued up to week 4, after which compaction slowly decreased. (B) The thickness of the samples decreased in the first 4 weeks, after which it rapidly increased to reach a constant thickness from week 6 on. Width decreased up to week 4, after which a small increase was observed towards week 8. (C) The surface area of the TE constructs decreased from week 2 to 4, increased from week 4 to 6 and then remained stable (D) Force was generated after 14 days and increased up to week 4, after which it remained stable. (E) Stress was generated after 14 days, increased up to week 4, followed by a decrease until week 6, after which it remained stable. 95

104 Chapter 6 96

105 Prolonged tissue culture reduces stress generation and retraction Tissue retraction and generated stress after release of constraints (constrained configuration) At release of constraints, the generated force (Figure 6.7A) gradually increased and reached a mean value of 16.2±1.0 mn after 30 minutes at week 4. At week 6, the generated force decreased to 7.9±1.5mN, while at week 8 it increased again to 12.7±2.1 mn. The generated stress after 30 minutes was 11.8±0.9 kpa at week 4 and decreased to 1.4±0.3 and 2.4±0.4 kpa at week 6 and 8, respectively (Figure 6.7B). During the first 30 minutes retraction (Figure 6.7C) occurred rapidly, reaching 30 40% of the total retraction after 24 hours. After 30 minutes, no differences were observed between retraction at weeks 4, 6 and 8 (12.8±1.8%, 9.3±1.0% and 10.6±1.2%, respectively). However, after 24 hours, retraction was more pronounced at week 4 (44.0±3.7%) compared to week 6 and 8 (29.2±2.8% and 26.1±2.2%, respectively). In the semiconstrained TE constructs, retraction at week 8 also occurred very rapidly in the first 30 minutes reaching a mean value of 8.7±1.3%. Subsequent changes were less dramatic leading to a retraction value of 18.4±3.0% after 24 hours (Figure 6.7C). Figure 6.7: Generated force (A), stress (B) and retraction (C) at week 4, 6 and 8 after release of constraints. (A) Generated force, 30 minutes after release of constraints, was lower at week 6 compared to week 4, and generated force at week 8 was in between week 4 and 6. (B) Generated stress was higher at week 4 compared to week 6 and 8. (C) After release of constraints, retraction occurred fast in the first 30 minutes after which it slowly increased. After 24 hours, retraction was higher at week 4 than at week 6 and 8 (constrained and semi constrained). 97

106 Chapter Variations and correlations within the experiment and configurations An overview of both runs has been provided in table 6.1. Overall, trends over time are similar with some parameters lower in the second run. Both runs have been normalized to statistically investigate differences between week 4, 6 and 8, and between the constrained and semi constrained configuration. Compaction slightly decreased after week 4, which became significant at week 8 (Figure 6.8A). The generated force was higher in the semi constrained samples, but in both configurations it remained constant between weeks 4, 6 and 8 (Figure 6.8B). Total stress (Figure 6.8C) was higher in the semi constrained than in the constrained configurations, while retraction at week 8 was similar in both configurations (Figure 6.8E). Both retraction and stress (constrained and semi constrained) decreased at weeks 6 and 8 compared to week 4. Figure 6.8: Normalized compaction (A), force (B), stress (C) and retraction (D) at week 4, 6 and 8 in constrained and semi constrained tissue culture. (A) Compaction decreased at week 8. (B) Total generated force was higher in the semi constrained configuration, but constant over time. (C and D) Both stress and retraction were less at week 6 and 8 compared to week 4. Generated stress was higher in the semi constrained samples than in the constrained ones. * denotes a significant difference compared to week 4. # denotes a significant difference of semi constrained culture compared to constrained culture at week 8. Single or double symbols indicate P<0.05 or P<0.001, respectively. 98