perfusion bioreactors for bone tissue engineering - a combined

Research Collection

Doctoral Thesis

Perfusion bioreactors for bone tissue engineeringA combined experimental and computational approach

Author(s): Vetsch, Jolanda R.

Publication Date: 2015

Permanent Link: https://doi.org/10.3929/ethz-a-010540254

Rights / License: In Copyright - Non-Commercial Use Permitted

This page was generated automatically upon download from the ETH Zurich Research Collection. For moreinformation please consult the Terms of use.

ETH Library

https://doi.org/10.3929/ethz-a-010540254

http://rightsstatements.org/page/InC-NC/1.0/

https://www.research-collection.ethz.ch

https://www.research-collection.ethz.ch/terms-of-use

DISS. ETH NO. 22998

Perfusion bioreactors for bone tissueengineering - A combined

experimental and computationalapproach

A thesis submitted to attain the degree of

DOCTOR OF SCIENCES of ETH ZURICH(Dr. sc. ETH Zurich)

presented by

Jolanda Rita Vetsch

MSc, ETH Zurich, Switzerland

born on 27th April, 1987

citizen of Grabs SG, Switzerland

accepted on the recommendation of

Prof. Dr. Ralph Müller, examinerProf. Dr. Sandra Hofmann, co-examiner

2015

“There must be a beginning of any great matter,

but the continuing unto the end until it be

thoroughly finished yields the true glory.”

Sir Francis Drake, 1587

Table of Contents

Acknowledgements iii

Summary vii

Zusammenfassung xi

1 Introduction 11.1 Thesis motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31.2 Specific aims . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51.3 Outline of the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 Background 112.1 Computational approaches . . . . . . . . . . . . . . . . . . . . . . . . 13

3 Development of perfusion bioreactor system 473.1 Perfusion bioreactor design . . . . . . . . . . . . . . . . . . . . . . . . 493.2 Static culture conditions - fetal bovine serum . . . . . . . . . . . . . . 573.3 Dynamic culture conditions - flow velocity . . . . . . . . . . . . . . . 84

4 Implementation of perfusion bioreactor system 1054.1 Influence of curvature on mineralized matrix . . . . . . . . . . . . . . 107

5 Synthesis 131

Curriculum Vitae 143

i

Acknowledgements

The research presented in this thesis would not have been possible without thesupport of a number of people to whom I’m greatly indebted.

First of all, I would like to thank Prof. Dr. Ralph Müller for giving me theopportunity to perform this work at the Institute for Biomechanics. His persistententhusiasm for research has been highly motivating for me. It was very importantfor me to feel his support, especially during times when I saw very little progress.I very much appreciate the frank discussions we had and his constructive criticismpushing me to perform better not only as a scientist but also as a person. I willalways be grateful for the lessons I learnt under his supervision.

Second, I am very deeply indebted to my co-referee and direct supervisor Prof.Dr. Sandra Hofmann. I’m still impressed by her vast scientific experience, notonly in tissue engineering, but also in many other scientific fields. I very muchappreciate that she gave me the freedom to decide on which route to take andbeing there for me when I thought that I was starting to get lost. Even aftermoving to the Netherlands she was always available to give quick feedback if it wasnecessary. I cannot emphasize enough the importance of her supervision, teachingme everything that I needed to accomplish my PhD. Sometimes, when I was workingon a manuscript reviewed by her the number of comments was almost overwhelming,but at the same time I realized that the reviews of my student’s manuscripts lookexactly the same. I am grateful that we have been able to get along very well not onlyin work but also in personal life while running through the woods, at conferences,during lunch breaks or when visiting her in the Netherlands.

I would like to thank all current and past members of the skeletal tissue engi-neering and the bone biomechanics group. Particularly I want to mention Dr. SilkeWüst, who was not only my officemate for almost 3 years, but is also linked to someprecious memories like ski events, conferences and lunch breaks at the Werdinsel;my current officemates Dr. Marina Rubert and Dr. Carly Tayler who gave meimportant scientific input and inspired me with new ideas, but also supported memorally when things did not go as well as planned; former and current membersof the HPI crew: Thomas Steiner, Dr. Luc Nimeskern, Dr. Kathryn Stok, Dr.

iii

Acknowledgements

Benjamin Thimm, Felicitas Flohr, Ariane Scheuren, Iina Lehtoviita, Dr. AndreasTrüssel, (soon to be Dr.) Sandro Badilatti, Dr. Alexander Zwahlen, Dr. AlinaLevchuk, Marios Georgiadis and Michele Casanova: it was a pleasure working withyou in such a supporting environment.

Furthermore, I would like to acknowledge past and present members of the Insti-tute for Biomechanics. Special thanks go to Peter Schwilch and Marco Hitz. Theyhave been a great help advising me with their broad technical background. Espe-cially, I want to thank Peter for his untiring patience reading my CAD drawings,but also for taking me back to the ’real world’ during lunch breaks. Many thanks goto Prof. Dr. Stephen Ferguson, Dr. Davide Ruffoni and Dr. René Widmer Soykawho supported me with important scientific input during my thesis. I also want tothank Dr. Gisela Kuhn for her advice on μCT and animal experiment related issues.Next, I would like to thank Sandro Badilatti, Dr. Kathryn Stok, Duncan Betts andBryce Besler who always had a ready ear for my problems with the VMS system.

Moreover, I want to thank all the students and apprentices that contributed tothis work. Specifically, I want to thank Gratianne Vaisson, Rahel Meister, Steve Ho,Manuela Estermann, Angela Mühlenbroich, Lukas Frey, Jenny Wu, Fabian Gsponerand Luca Plan for their efforts and motivation to be part of this project. A spe-cial thank goes to Samantha Paulsen, who spent a full year working with me as aWhittaker fellow. Her contributions to this work have been of high value. I highlyappreciate her motivation and eager work attitude as well as her cheerful habit ofmind. I would always work with her again if I had the opportunity!

There were also a lot of people from outside the Institute supporting this thesisin one way or another. In particular, I would like to thank Dr. Dirk Mohn whohelped me with FTIR measurements and evaluation, granting me access to theirautoclave and microscope as well as for the various fruitful discussions accompaniedby a cup of the best coffee served on campus, or during longjogs through the woods.Additionally, I would like to acknowledge Nora Hild and Roland Fuhrer for theirsupportive help.

During the past years, I have worked with several collaborators: Marianne Som-mer, Dr. Ali Mirsaidi, Dr. Emilie Zermatten, Dr. Cheryl Rahman, PD Dr. PeterRichards and Prof. Dr. Charles Sfeir. Many thanks for your trust and the excellentcollaborations.

Many of the above named work relationships turned into friendships which Iappreciate a lot. Surely there will be a lot people leaving traces in my future life,thank you very much!

This work would not have been possible without the understanding and support

iv

Acknowledgements

of all my friends and my family. I apologize for all the missed and postponedappointments or for being late, because I couldn’t finish on time in the lab. Iwould like to thank my parents, who always supported me in my decisions since mychildhood and during my entire education. My brother Andreas, with whom I couldtalk about every topic openly and my brother Christian: soon you are officiallyallowed to call me Dr. Vetsch.

The biggest thanks go to my soon-to-be husband Lukas. Words cannot expressmy deep gratitude for your unconditional love, faithful support and cheering upsespecially when I was living through challenging times. I am incredibly lucky tohave you on my side and I am very much looking forward to what the future willbring.

Finally, I would like to acknowledge financial support from the European Union’sSeventh Framework Programme (FP/2007–2013) under Grant Agreement n. 262948.

September 2015Jolanda R. Vetsch

v

Summary

Conventional bone replacement strategies exhibit various disadvantages like lim-ited supply, post-operative complications or donor site morbidity. The replacementof large bone defects due to trauma or congenital diseases is still a major issue.Bone tissue engineering (BTE) tries to overcome these drawbacks aiming at grow-ing functional cell-matrix constructs by the combination of scaffold materials, cellsand environmental cues, like growth factors or mechanical loading.

To generate functional cell-matrix constructs in-vitro various bioreactor designshave been proposed. Bioreactors are defined devices that house tissue-engineeredconstructs under controlled conditions providing a high degree of reproducibility.Different dynamic bioreactor designs have been proposed for BTE applications.Compression and perfusion bioreactors are the most frequently used designs in BTE.They have been shown to enhance in-vitro osteogenesis in BTE cultures by the ap-plication of mechanical forces.

In-vivo, mechanical loading plays an important role in healthy bone as well asduring fracture healing. Bone is constantly remodelled adapting its histologicalstructure to changes in long-term mechanical loading. In healthy bone, osteocytesare the mechanosensitive cells orchestrating bone remodelling based on the inten-sity of the mechanical signal. Osteocytes are sitting within the lacuno-canalicularnetwork that is filled with interstitial fluid. Due to bone matrix deformations theinterstitial fluid flow exerts shear stresses (SS) sensed by the osteocytes. Duringfracture healing cells within the repair tissue are exposed to SS as well. Cellularresponses have been shown to be influenced by the loading regime present withinregeneration constructs. Precursor cells like human mesenchymal stem cells (hM-SCs) play an important role in fracture healing and have been shown to be a highlysuitable cell source for clinical applications, because they can be easily harvestedwith minimal donor site morbidity and exhibit a high proliferation potential in-vitro.The mechanosensitivity of hMSCs has been proven in different studies showing os-teogenic differentiation of hMSCs when subjected to perfusion-induced SS. Com-pared to other bioreactor designs perfusion bioreactors are thought to resemble thein-vivo loading of bone the closest.

vii

Summary

Perfusion bioreactors have been shown to positively influence various cell types inBTE cultures, but some studies reported contradictory results leading to the assump-tion that numerous influencing parameters of BTE cultures are still not completelyunderstood. The implementation of micro-computed tomography (μCT) is able toadd high value to BTE cultures. Quantitative three-dimensional (3-D) evaluationof tissue morphology or scaffold geometry can be performed without destroying thesample leaving it intact for prolonged cell culture. The combination of computationalsimulations with μCT is a powerful tool to calculate various aspects of cell cultureslike tissue growth or mechanical environment. Computational simulations are ableto help drawing causal relations between influencing parameters and experimentaloutcomes, ultimately leading to a better understanding of biological processes inBTE cultures.

This thesis aimed at establishing a perfusion bioreactor system with optimizedculture conditions for BTE applications. The thesis was divided into three aims:(i) the development of a perfusion bioreactor design for monitoring 3-D mineralizedtissue formation using μCT. (ii) the optimization of culture conditions for BTEcultures using hMSCs on silk fibroin (SF) scaffolds, and (iii) the application of theperfusion bioreactor system developed using the optimized culture conditions toinvestigate the influence of curvature on mineralized tissue formation. In a firststep, the design of our in-house designed perfusion bioreactor was improved usingcomputational fluid dynamics. It was shown that the original bioreactor design led toinhomogeneous and non-uniform velocity fields within the bioreactor and the scaffoldespecially at higher flow rates. Three different approaches were able to preventthe formation of inhomogeneous velocity fields: (i) increasing the height of thebioreactor, (ii) increasing the diameter of the inlet and outlet, and (iii) the insertionof flow conditioners. Increasing the height of the bioreactor was not feasible due tosize limitations given by the μCT machine and the application of flow conditionersled to massive air inclusions inside the bioreactor that could not be eliminated oncepresent. Therefore, the original design of the perfusion bioreactor was adapted byincreasing the diameter of the inlet and outlet to 5mm which led to homogeneousvelocity fields in the bioreactor as well as in the scaffold.

Next, culture conditions, namely culture medium supplementation with fetalbovine serum (FBS) and the mechanical loading regime were optimized for theperfusion bioreactor design developed. It was shown that different FBS types in-duced spontaneous mineralization on acellular SF scaffolds. The mineralization oncell-seeded SF scaffolds was shown to be cell-mediated only. Additionally, it wasshown that the mechanical loading regime applied influenced the cellular behaviour

viii

Summary

of hMSCs cultured on SF scaffolds. A lower flow rate led to proliferation of hMSCswhereas a higher flow rate led to differentiation and subsequent mineralized tis-sue formation by hMSCs. Based on these results suitable control group conditionsand a mechanical loading regime inducing mineralized tissue formation have beendetermined for future in-vitro experiments.

Finally, the perfusion bioreactor system developed using the optimized cultureconditions was successfully applied to investigate the influence of curvature on 3-Dmineralized tissue formation by hMSCs in SF scaffolds. SF scaffolds with differentchannels of different curvatures have been produced. The influence of curvature on 3-D mineralized tissue formation was investigated under static and loaded conditions.It could be shown that the mineralized tissue formation was dependent on curvatureand was additionally enhanced by mechanical loading. Interestingly, the morphologyof the mineralized tissue formed was highly dependent on the mechanical loadingcondition applied. Static samples exhibited cortical-like mineralized tissue structure,whereas loaded samples exhibited trabecular-like mineralized tissue structure. Theresults of the study presented suggest that the 3-D in-vitro model is not only able toshow the effects of curvature on mineralized tissue formation, but could be used infuture experiments as a model for circular critical size defects in cortical or trabecularbone.

In conclusion, a perfusion bioreactor system has been established by adapting theoriginal perfusion bioreactor design leading to a homogeneous mechanical environ-ment within the bioreactor and the scaffold. The knowledge gained from optimizingculture conditions was combined with the adapted bioreactor design to study theinfluence of curvature on 3-D mineralized tissue formation. The perfusion bioreactorsystem developed is believed to serve as a framework providing a defined mechanicalenvironment and defined culture conditions to investigate and understand causal re-lations between different influencing parameters and experimental outcomes in BTEapplications. Ultimately, the increased understanding could be used to guide thedevelopment and increase the quality of novel strategies for bone replacement.

ix

Zusammenfassung

Gängige Strategien für den Knochenersatz weisen verschiedene Nachteile wie lim-itierte Versorgung, post-operative Komplikationen oder Morbidität im Spender-bereich auf. Der Ersatz von grossen Knochendefekten, verursacht durch Trau-mata oder Erbkrankheiten, ist bis heute ein bedeutendes Problem. Die Züchtungvon Knochengewebe im Labor (Knochen Tissue-Engineering, KTE) versucht dieseNachteile zu überwinden. Das Ziel des KTEs ist funktionelle Zell-Matrix Kon-strukte, durch Kombination von verschiedenen Trägermaterialien (Scaffolds), Zellenund Umweltfaktoren, wie zum Beispiel Wachstumsfaktoren oder mechanische Be-lastung, herzustellen.

Verschiedene Bioreaktormodelle sind für die in-vitro Züchtung von funktionellenZell-Matrix Konstrukten vorgeschlagen worden. Bioreaktoren sind definierte Geräte,die eine Kultivierung von gezüchtetem Gewebe unter kontrollierten Bedingungen er-möglichen und einen hohen Grad an Reproduzierbarkeit aufweisen. Für das KTEwurden verschiedene dynamische Bioreaktormodelle entwickelt. Die am häufigstenverwendeten Modelle sind Kompressions- und Durchflussbioreaktoren. Es konntegezeigt werden, dass diese die in-vitro Knochenbildung durch Anwendung von mech-anischen Kräften fördern.

Mechanische Belastung spielt in gesunden Knochen sowie auch während derFrakturheilung eine wichtige Rolle. Gesunde Knochen werden ununterbrochenremodelliert um ihre Struktur an dauerhafte Veränderungen der mechanischenBeanspruchungen anzupassen. Osteozyten sind mechanosensorische Zellen, welchedie Knochenremodellierung basierend auf der von ihnen gespürten Intensitätder mechanischen Signale regulieren. Osteozyten liegen eingebettet im lakuno-kanalikulären Netzwerk des Knochens, dessen Zwischenräume mit Flüssigkeit gefülltsind. Deformationen der Knochenmatrix führen zu einem Flüssigkeitsfluss, derScherkräfte generiert, die von den Osteozyten gespürt werden. Auch während derFrakturheilung sind die Zellen im Reparaturgewebe Scherkräften ausgesetzt. Es kon-nte gezeigt werden, dass mechanische Belastung von Konstrukten für den Knoch-enersatz die zelluläre Antwort beeinflussen kann. Vorläuferzellen wie zum Beispielhumane mesenchymale Stammzellen (hMSCs) spielen eine wichtige Rolle während

xi

Zusammenfassung

der Frakturheilung. hMSCs sind für klinische Anwendungen sehr gut geeignet, dasie einfach gewonnen werden können und ein hohes Potential für die in-vitro Ver-mehrung aufweisen. hMSCs sind wie Osteozyten mechanosensorische Zellen. Ver-schiedene Studien haben gezeigt, dass hMSCs, die mit Scherkräften mechanischbelastet wurden, in Richtung der osteogenen Zellfamilie differenzieren. Im Vergleichzu anderen Bioreaktorsystemen ist der Durchflussbioreaktor das Modell, welches diemechanische Belastung von Knochen in-vivo am ähnlichsten wiedergibt.

Die Anwendung von Durchflussbioreaktoren für das KTE hat verschiedeneZelltypen positiv beeinflusst. Es wurden aber auch widersprüchliche Resultatebeobachtet, was zu der Annahme führt, dass die Einflussfaktoren von Kulturenfür das KTE noch nicht klar verstanden werden. Die Implementierung von Mikro-Computertomographie (μCT) ist von hohem Wert für das KTE. Mithilfe von μCTkann die dreidimensionale (3-D) Gewebemorphologie oder die Geometrie von Scaf-folds quantitativ bestimmt werden ohne das Sample zu zerstören, um es für län-gere Zellkulturen weiter zu verwenden. Die Kombination von μCT und com-putergestützten Simulationsverfahren ist ein leistungsfähiges Instrument um ver-schiedene Aspekte von Zellkulturen, wie das Gewebewachstum oder die mechanischeUmgebung, zu berechnen. Computergestützte Simulationsverfahren helfen kausaleBeziehungen zwischen Einflussfaktoren und experimentellen Ergebnissen zu verste-hen. Dies kann letztendlich zu einem verbesserten Verständnis von biologischenProzessen in Kulturen für das KTE führen.

Diese Doktorarbeit hatte das Ziel ein Durchflussbioreaktor-System mit opti-mierten Kultivierungsbedingungen für das KTE zu etablieren. Die Arbeit wurdein drei verschiedene Ziele gegliedert: (i) Die Entwicklung der Konstruktion einesDurchflussbioreaktors für die Verfolgung des 3-D Wachstums von mineralisiertemGewebe mithilfe von μCT, (ii) die Optimierung von Kultivierungsbedingungen fürdas KTE durch hMSCs auf Seidenfibroin (SF) Scaffolds, und (iii) die Anwendungdes Durchflussbioreaktor-Systems mit den optimierten Kultivierungsbedingungenum den Einfluss von Kurvatur auf das 3-D Wachstum von mineralisiertem Gewebezu untersuchen.

In einem ersten Schritt wurde die Konstruktion unseres selbst-entwickelten Durch-flussbioreaktors mithilfe von computergestützter Strömungsdynamik verbessert.Die ursprüngliche Konstruktion führte zu inhomogenen und ungleichmässigenFlussgeschwindigkeitsfeldern im Bioreaktor und im Scaffold vor allem bei höherenFlussraten. Drei verschiedene Ansätze konnten die Bildung von inhomogenenFlussgeschwindigkeitsfeldern verhindern: (i) Eine Erhöhung des Bioreaktors, (ii)eine Vergrösserung des Durchmessers des Ein- und Auslasses und (iii) das Einsetzen

xii

Zusammenfassung

von Strömungsgleichrichtern. Die Erhöhung des Bioreaktors war nicht realisier-bar, da die maximale Höhe durch die μCT-Maschine begrenzt ist. Das Einsetzenvon Strömungsgleichrichtern führte zu massiven Lufteinschlüssen im Bioreaktor, dienicht mehr eliminiert werden konnten. Der Durchmesser des Ein- und Auslassesder ursprünglichen Konstruktion des Durchflussbioreaktors wurde schlussendlich auf5mm vergrössert, was zu homogenen Flussgeschwindigkeitsfeldern im Bioreaktor undim Scaffold führte.

Als nächstes wurden die Kultivierungsbedingungen, genauer gesagt die Ergänzungdes Kulturmediums mit fötalem bovinen Serum (FBS) und der mechanische Belas-tungsmodus, optimiert. Verschiedene Typen von FBS führten zu spontaner Min-eralisierung in azellulären SF Scaffolds. Es konnte gezeigt werden, dass die Miner-alisierung, die auf zellulären SF Scaffolds beobachtet wurde, zellvermittelt gebildetwurde. Der mechanische Belastungsmodus hat das zelluläre Verhalten von hMSCsauf SF Scaffolds direkt beeinflusst. Tiefere Flussraten führten zur Vermehrung derhMSCs, wohingegen höhere Flussraten zur Differenzierung und anschliessender Bil-dung von mineralisiertem Gewebe durch die hMSCs führten. Basierend auf diesenResultaten wurden geeignete Konditionen für Kontrollgruppen und einen mechanis-chen Belastungsmodus, der die Bildung von mineralisiertem Gewebe induziert, fürzukünftige in-vitro Experimente festgelegt.

Das entwickelte Durchflussbioreaktor-System wurde schliesslich erfolgreich für dieUntersuchung des Einflusses von Kurvatur auf das 3-D Wachstum von mineral-isiertem Gewebe durch hMSCs auf SF Scaffolds angewandt. SF Scaffolds wurdenmit verschiedenen Kanälen, die verschiedene Kurvaturen repräsentieren, produziert.Der Einfluss der Kurvatur wurde unter statischen und belasteten Konditionen unter-sucht. Das Wachstum des mineralisierten Gewebes war abhängig von der Kurvaturund wurde durch die Anwendung von mechanischer Belastung zusätzlich beeinflusst.Die Morphologie des mineralisierten Gewebes war interessanterweise sehr stark vonder mechanischen Belastung abhängig. Unter statischen Bedingungen zeigte dasmineralisierte Gewebe eine kortikale Struktur, während unter belastenden Bedin-gungen eine trabekuläre Struktur beobachtet wurde. Die Resultate dieser Studieweisen darauf hin, dass das 3-D in-vitro Modell nicht nur die Effekte von Kurvaturauf das Wachstum von mineralisiertem Gewebe aufzeigen kann, sondern in zukün-ftigen Studien auch als Modell für zirkuläre Knochendefekte kritischer Grösse inkortikalem oder trabekulärem Knochen angewendet werden kann.

Abschliessend kann festgestellt werden, dass durch die Anpassung der ur-sprünglichen Konstruktion des Durchflussbioreaktors ein Durchflussbioreaktor-System mit homogener mechanischer Umgebung im Bioreaktor und im Scaffold

xiii

Zusammenfassung

entwickelt wurde. Das während der Optimierung der Kultivierungsbedingungenerworbene Wissen wurde mit der angepassten Konstruktion des Durchflussbioreak-tors kombiniert um den Einfluss von Kurvatur auf das 3-D Wachstum von mineral-isiertem Gewebe zu untersuchen. Das entwickelte Durchflussbioreaktor-System sollin Zukunft als Framework dienen, das eine definierte mechanische Umgebung unddefinierte Kultivierungsbedingungen unterstützt, um kausale Zusammenhänge zwis-chen Einflussfaktoren und experimentellen Ergebnissen von Kulturen für das KTEzu verstehen. Das verbesserte Verständnis könnte schlussendlich für die Entwicklungund die Verbesserung der Qualität von neuartigen Strategien für den Knochenersatzangewendet werden.

xiv

Chapter 1

Introduction

1

1.1 Thesis motivation

1.1 Thesis motivation

Conventional strategies for bone replacement in orthopedics include patient-derived(autologous) or donor-derived (allogeneic, xenogeneic) grafts. These approacheshave several disadvantages like limited supply, associated post-operative complica-tions or, in the case of autologous grafts, donor site morbidity [1]. Bone has theunique property to heal itself, but its repair capacity decreases with increasing bonedefect size or the amount of interfragmentary movement [2]. An impaired repair ca-pacity of bone could ultimately lead to delayed unions, malunions or nonunions [1].To this day, the treatment of large bone defects is still a major issue. Bone tissue en-gineering (BTE) tries to overcome the above named drawbacks of conventional bonereplacement strategies. The overall aim of BTE is to grow a functional cell-matrixconstruct. For this purpose different scaffold materials supporting cell attachmentand maintenance of cell function, a high number of osteoprogenitor cells and suit-able growth factors can be used in a variety of different combinations to enhancecell proliferation and differentiation [2, 3].

To generate functional cell-matrix constructs in-vitro specialized bioreactors havebeen proposed. Bioreactors are defined as devices in which tissue-engineered con-structs develop under closely monitored and tightly controlled conditions in a sterileenvironment [4]. The advantages of bioreactors are a high degree of reproducibility,control and automation [5]. Different dynamic bioreactor designs have been pro-posed to improve cell seeding and mass transport or to apply mechanical loading inBTE cultures. The two most prominent bioreactor types used for BTE applicationsare compression and perfusion bioreactors. Both bioreactor types have been shownto enhance in-vitro osteogenesis [4].

In-vivo, mechanical loading plays a crucial role during bone remodeling as wellas during fracture repair. Healthy bone tissue is constantly undergoing remodelingprocesses allowing the bone to adapt its histological structure to changes in long-termmechanical loading [6]. The osteocyte is considered to be the mechanosensitive cell inhealthy bone [7,8]. Sitting in the lacuno-canalicular network of bone osteocytes areable to sense their mechanical environment via interstitial fluid flow [9–11]. Drivenby bone matrix deformations the interstitial fluid flow exerts shear stresses (SS) onosteocytes, which translate the physical force into corresponding intracellular signals[12]. Compared with healthy bone, cells within the repair tissue of a fracture areloaded by SS, too. Cellular responses and tissue differentiation patterns have beenshown to be influenced by the loading regime applied on bone regeneration constructs[13]. Precursor cells like human mesenchymal stem cells (hMSCs) are known to playa major role in early fracture healing. The mechanosensitivity of hMSCs has been

3

Chapter 1 Introduction

proven by the application of perfusion-induced SS leading to a differentiation ofthe hMSCs towards the osteogenic lineage [14, 15]. For clinical BTE applications,hMSCs are a highly suitable cell source due to their ease of harvesting from varioustissues with minimal donor site morbidity and large proliferation potential, leadingto a high efficiency on cell yield in-vitro. Another advantage of hMSCs is that theycan be harvested autologously leading to very low incidence of rejection due to thedonor and recipient being the same individual [16, 17].

Based on these observations, perfusion bioreactors are considered to translatethe concept of fluid flow induced mechanical loading in-vivo to in-vitro cell culturemodels the closest [18]. Perfusion bioreactors have been shown to enhance osteogenicdifferentiation, increase levels of osteogenic markers and enhance mineralized matrixdeposition of various cell types [14, 19, 20]. Dynamic BTE cultures might thereforebe a crucial means to improve the quality and functionality of engineered constructs.

Despite the evidence that various cell types are positively influenced by the ap-plication of perfusion bioreactors some studies also reported cell apoptosis or cellproliferation rather than differentiation [14,19]. These results suggest that there arenumerous influencing parameters of BTE cultures that are still not clearly under-stood.

Micro-computed tomography (μCT) is a quantitative, non-invasive three-dimensional (3-D) imaging technique that is able to add high value to BTE cultures.The evolution of mineralized tissue can be monitored longitudinally and various mor-phological parameters can be derived in analogy to bone [21,22]. Compared to otherstrategies used to image tissue morphology, for instance histology, μCT scans can beperformed without destroying the sample leaving it intact for prolonged cell culture.In addition, 3-D scaffold geometries can be obtained at high resolutions allowingdirect comparisons between scaffold structure and cell response. The combinationof computational simulations with μCT is a powerful means to investigate variousaspects of cell cultures like mechanical loading conditions [23], tissue growth [24], so-lute concentration [25] and many more. Computational simulations are able to helpdrawing causal relations between influencing parameters and experimental outcomesof BTE cultures. Nevertheless, it is almost impossible considering every parame-ter of BTE cultures in computational simulations as well as in experiments. It isnecessary to examine and to understand single influencing parameters in simplifiedmodels at first. This could subsequently lead to an improved general understandingof biological processes in BTE cultures.

With this in mind, we proposed to establish a perfusion bioreactor system forBTE cultures with the possibility for μCT application based on experimental and

4

1.2 Specific aims

computational results. The developed system would serve as framework providinga defined mechanical environment and optimized culture conditions for dynamicBTE cultures. The evaluation of causal relations between influencing parametersand experimental outcomes is assumed to improve the understanding of biologicalprocesses in BTE cultures, ultimately leading to a higher quality of engineered bonetissue.

1.2 Specific aims

The overall aim of this thesis was to establish a perfusion bioreactor system forthe monitoring of 3-D mineralized tissue formation with μCT and to optimize itsculture conditions for BTE applications based on experimental and computationaldata. The established system was applied using the optimized culture conditions toinvestigate the influence of curvature on 3-D mineralized tissue formation of hMSCscultured on silk fibroin (SF) scaffolds.

Specifically, the following three aims have been defined:

Aim 1: Development of a perfusion bioreactor design for monitoring 3-D mineral-ized tissue formation using μCT.

Aim 2: Optimization of culture conditions for BTE cultures using hMSCs culturedon SF scaffolds.

Aim 3: Application of the perfusion bioreactor system using the optimized cultureconditions to investigate the influence of curvature on 3-D mineralized tissueformation of hMSCs on SF scaffolds.

1.3 Outline of the thesis

The thesis is structured into 5 chapters.

Chapter 1 presents the motivation, the specific aims and the outline of this thesis.

Chapter 2 provides a background about the evolution of simulation techniques ap-plied for dynamic BTE cultures in bioreactors. Different computational ap-proaches simulating the mechanical or biological environment in the four mostprominent bioreactor types used for dynamic BTE are displayed. The potentialof combining computational simulations with quantitative imaging techniques

5


like μCT is specified and it is outlined how these techniques could be usedin the future to improve the understanding of the development of a tissue-engineered construct in response to mechanical loading in a bioreactor.

Chapter 3 describes the development of a perfusion bioreactor system for dynamicBTE applications with the possibility to monitor 3-D mineralized tissue for-mation using μCT. The first section focuses on the improvement of the designof our in-house designed perfusion bioreactor. Different bioreactor geometrieswere designed and simulated to investigate the effect of bioreactor design onthe flow fields in the bioreactor and the scaffold. In the second part of thechapter, the optimization of culture conditions for static BTE cultures is pre-sented. The consequence of medium supplementation with different animalsera is emphasized with respect to 3-D mineralized tissue formation on SFscaffolds. The third section describes the optimization of culture conditionsfor dynamic BTE cultures. The effects of two different flow rates on the be-havior of hMSCs cultured on SF scaffolds in the perfusion bioreactor designdeveloped were investigated. Computational modeling was used to investigatethe mechanical environment within the scaffold. Taken togehter, a bioreactordesign was developed leading to a homogeneous mechanical environment in thebioreactor and the scaffold. Based on the evaluation of the effect of the ani-mal sera, one serum was chosen to provide optimal culture medium conditionsfor perfusion cultures and a mechanical loading regime inducing mineralizedtissue formation was defined. The knowledge gained was applied in future ex-periments to control cellular behavior within the perfusion bioreactor systemdeveloped.

Chapter 4 is concerned with the application of the developed perfusion bioreactorsystem with the optimized culture conditions. In particular, the system de-veloped was applied to investigate the effect of curvature on 3-D mineralizedtissue formation. SF scaffold were produced with three different channels, rep-resenting three different curvatures. It was shown that 3-D mineralized tissueformation is directly linked to curvature. The combination of the perfusionbioreactor system, the optimized dynamic culture conditions and the scaffoldgeometry developed for the experiment led to a novel model for investigatingcircular critical size defects in-vitro for trabecular and cortical bone.

Chapter 5 contains the synthesis of the presented thesis including key findings,contribution to the research in this field, limitations and an outlook on futurework.

6

1.2 Specific aims

References

[1] C. G. Finkemeier. Bone-grafting and bone-graft substitutes. J Bone Joint SurgAm, 84-A(3):454–64, 2002.

[2] A. Liedert, D. Kaspar, P. Augat, A. Ignatius, L. Claes. Mechanobiology ofBone Tissue and Bone Cells, pages 418–433. Academia Publishing House Ltd.,Moscow, Russia, 2005.

[3] F. R. Rose and R. O. Oreffo. Bone tissue engineering: hope vs hype. BiochemBioph Res Co, 292(1):1–7, 2002.

[4] R. I. Abousleiman and V. I. Sikavitsas. Bioreactors for tissues of the muscu-loskeletal system. Adv Exp Med Biol, 585:243–59, 2006.

[5] I. Martin, D. Wendt, M. Heberer. The role of bioreactors in tissue engineering.Trends Biotechnol, 22(2):80–6, 2004.

[6] S. C. Cowin and D. H. Hegedus. Bone remodeling I: theory of adaptive elasticity.J Elast, 6(3):313–326, 1976.

[7] L. F. Bonewald. Mechanosensation and transduction in osteocytes. BonekeyOsteovision, 3(10):7–15, 2006.

[8] C. R. Jacobs, S. Temiyasathit, A. B. Castillo. Osteocyte mechanobiology andpericellular mechanics. Annu Rev Biomed Eng, 12:369–400, 2010.

[9] E. H. Burger and J. Klein-Nulend. Mechanotransduction in bone–role of thelacuno-canalicular network. FASEB J, 13 Suppl:S101–12, 1999.

[10] S. P. Fritton and S. Weinbaum. Fluid and solute transport in bone: Flow-induced mechanotransduction. Annu Rev Fluid Mech, 41:347–374, 2009.

[11] J. Klein-Nulend, R. G. Bacabac, A. D. Bakker. Mechanical loading and howit affects bone cells: the role of the osteocyte cytoskeleton in maintaining ourskeleton. Eur Cell Mater, 24:278–91, 2012.

[12] A. C. Allori, A. M. Sailon, J. H. Pan, S. M. Warren. Biological basis of boneformation, remodeling, and repair-part III: biomechanical forces. Tissue EngPart B Rev, 14(3):285–93, 2008.

[13] M. Liebschner and M. Wettergreen. Topics in tissue engineering, chapter Op-timization of Bone Scaffold Engineering for Load Bearing Applications. 2003.

7


[14] F. Zhao, R. Chella, T. Ma. Effects of shear stress on 3-D human mesenchymalstem cell construct development in a perfusion bioreactor system: Experimentsand hydrodynamic modeling. Biotechnol Bioeng, 96(3):584–95, 2007.

[15] W. L. Grayson et al. Optimizing the medium perfusion rate in bone tissueengineering bioreactors. Biotechnol Bioeng, 108(5):1159–70, 2011.

[16] M. Mullender et al. Mechanotransduction of bone cells in vitro: mechanobiologyof bone tissue. Med Biol Eng Comput, 42(1):14–21, 2004.

[17] D. Marolt, M. Knezevic, G. V. Novakovic. Bone tissue engineering with humanstem cells. Stem Cell Res Ther, 1(2):10, 2010.

[18] R. J. McCoy and F. J. O’Brien. Influence of shear stress in perfusion bioreactorcultures for the development of three-dimensional bone tissue constructs: areview. Tissue Eng Part B Rev, 16(6):587–601, 2010.

[19] S. H. Cartmell, B. D. Porter, A. J. Garcia, R. E. Guldberg. Effects of mediumperfusion rate on cell-seeded three-dimensional bone constructs in vitro. TissueEng, 9(6):1197–203, 2003.

[20] H. L. Holtorf, J. A. Jansen, A. G. Mikos. Flow perfusion culture inducesthe osteoblastic differentiation of marrow stroma cell-scaffold constructs in theabsence of dexamethasone. J Biomed Mater Res A, 72(3):326–34, 2005.

[21] H. Hagenmuller et al. Non-invasive time-lapsed monitoring and quantificationof engineered bone-like tissue. Ann Biomed Eng, 35(10):1657–67, 2007.

[22] T. Hildebrand, A. Laib, R. Muller, J. Dequeker, P. Ruegsegger. Direct three-dimensional morphometric analysis of human cancellous bone: microstruc-tural data from spine, femur, iliac crest, and calcaneus. J Bone Miner Res,14(7):1167–74, 1999.

[23] F. Boschetti, M. T. Raimondi, F. Migliavacca, G. Dubini. Prediction of themicro-fluid dynamic environment imposed to three-dimensional engineered cellsystems in bioreactors. J Biomech, 39(3):418–25, 2006.

[24] F. Galbusera, M. Cioffi, M. T. Raimondi, R. Pietrabissa. Computational mod-eling of combined cell population dynamics and oxygen transport in engineeredtissue subject to interstitial perfusion. Comput Methods Biomech Biomed En-gin, 10(4):279–87, 2007.

8

1.2 Specific aims

[25] S. Truscello, J. Schrooten, H. Van Oosterwyck. A computational tool for theupscaling of regular scaffolds during in vitro perfusion culture. Tissue Eng PartC Methods, 17(6):619–630, 2011.

9

Chapter 2

Background

11

2.1 Computational approaches

2.1 Computational approaches for dynamic bone

tissue engineering

The evolution of simulation techniques for dynamic bone tissue engineering in biore-actors

Jolanda Rita Vetsch1, Ralph Müller1, Sandra Hofmann1

1Institute for Biomechanics, ETH Zurich, 8093 Zurich, Switzerland

published in:Journal of Tissue Engineering and Regenerative Medicine2013, epubPostprint version according to publisher copyright policy.

Abstract:Bone tissue engineering aims to overcome the drawbacks of current bone regener-ation techniques in orthopaedics. Bioreactors are widely used in the field of bonetissue engineering, as they help support efficient nutrition of cultured cells withthe possible combination of applying mechanical stimuli. Beneficial influencingparameters of in-vitro cultures are difficult to find and are mostly determined bytrial and error, which is associated with significant time and money spent. Math-ematical simulations can support the finding of optimal parameters. Simulationshave evolved over the last 20 years from simple analytical models to complex anddetailed computational models. They allow researchers to simulate the mechanicalas well as the biological environment experienced by cells seeded on scaffolds ina bioreactor. Based on the simulation results, it is possible to give recommen-dations about specific parameters for bone bioreactor cultures, such as scaffoldgeometries, scaffold mechanical properties, the level of applied mechanical loadingor nutrient concentrations. This article reviews the evolution in simulating variousaspects of dynamic bone culture in bioreactors and reveals future research directions.

Keywords:dynamic tissue engineering, bone, bioreactor, simulation, mechanical stimuli, scaf-fold

13

Chapter 2 Background

2.1.1 Introduction

Bone tissue engineering combines the principles of engineering and life sciences toovercome drawbacks of traditional bone regeneration techniques used in orthopaedics[1]. The structural and mechanical characteristics of a tissue-engineered constructare intended to mimic the natural tissue as closely as possible and determine itssuccess in clinical applications. The first strategies in three-dimensional (3-D) bonetissue engineering consisted of static cultures, in which bone cells were seeded on ascaffold and placed in a well-plate for a defined period of time. This culture strategyled to different drawbacks: the cells tended to concentrate at the periphery of thescaffold, causing poor nutrient and waste exchange in the middle of the scaffold [2].This further led to cell necrosis in the centre of the scaffold. Dynamic bioreactorscould help prevent such problems in cell culture.

Bioreactors are defined as devices that enable a closely monitored and tightlycontrolled environment to allow biological and biochemical processes to develop.Bioreactors provide a high degree of reproducibility, control and automation, whichis favourable for specific experimental processes [3]. Additionally, some bioreactorsare able to apply physical stimuli, such as compression or shear stress, to the con-structs. It was shown in-vitro that mechanical stimulation improved cell behaviourand structure of engineered bone tissue using osteoblasts and stem cells [4–6].

Despite the numerous bioreactor designs, device parameters leading to improvedreproducibility in bone tissue cultures have not yet been determined systematically,possibly because, until now, the parameters have been chosen by trial-and-error-approaches. A possible strategy to determine the effect of parameters influencingtissue-engineered outcomes is to simulate tissue-engineering systems. Simulationsare able to compute stress and strain distributions, fluid shear stresses and veloc-ities, bone ingrowth and several other aspects of bone tissue-engineering cultures,depending on the scaffold’s properties. These simulations can then be comparedwith the obtained in-vitro results and parametric studies can indicate which factorshave a significant effect on the engineered output. The first simulations of bonetissue engineering in bioreactors were performed in the early 1990s. Since then, thefield of simulations in bone tissue engineering has evolved vastly. In the last decadethe method of finite element (FE) modelling became more and more important, re-placing earlier, simpler models. Micro-computed tomography (μCT) is a commonlyapplied imaging technique to obtain the 3-D geometry of the simulated bioreactor-scaffold system, which then can be directly implemented into an FE model.

In this review we focus only on bone tissue engineering and the four most com-monly used bioreactors in dynamic bone tissue engineering: (a) rotating wall vessel

14


bioreactor; (b) spinner flask bioreactor; (c) compression bioreactor; and (d) perfu-sion bioreactor (Table 2.1.1). Hydrodynamic bioreactors will not be discussed, dueto the lack of simulation studies in these bioreactors. For the same reason, somespinner flask studies cited in this publication were primarily intended for cartilagetissue engineering, but their results are potentially applicable to bone tissue engi-neering. Given the volume of work in the field of mathematical modelling in dynamicbone tissue engineering, it is not possible to be fully comprehensive. This reviewtherefore aims to summarize past and current work and reveal future research direc-tions that may be most relevant to optimizing bone tissue engineering. Optimizingcell-seeding strategies will not be covered in this review.

2.1.2 Bioreactors for dynamic bone tissue engineering

The first dynamic bioreactors were developed in the early 1990s [7–10], about 5 yearsafter the evolution of the field of tissue engineering [11]. Dynamic bioreactors havebeen designed to overcome the drawbacks of static cultures, such as poor nutrientand waste exchange [12, 13]. The four most prevalent bioreactor designs for bonetissue engineering are described in the following sections.

Rotating wall vessel bioreactor

The rotating wall vessel bioreactor is composed of two concentric cylinders(Fig. 2.1.1A). The outer cylinder is capable of rotating and the space between thetwo cylinders is filled with culture medium [14]. Scaffolds are freely suspended in theculture medium and are subjected to dynamic laminar flow [15], leading to low shearstresses and high mass-transfer rates [16]. Rotating wall vessel bioreactors have beenprimarily used for culturing cartilage tissue in-vitro [15] and only a few studies onbone tissue engineering exist. It has been shown that rotating wall vessel bioreactorscan improve the osteogenic differentiation of cells, increase mineralized extracellularmatrix (ECM) production and enhance the distribution of cells throughout the 3-Dscaffolds. However, cell growth and mineralization were still limited to the outer sur-faces of the 3-D scaffolds, because internal diffusion limitations were not eliminatedor cell sheets encapsulated the scaffold. The transport of nutrients to the centre ofthe scaffold was still limited because the convective forces could not extend to theinterior of the scaffold [2, 17].

15

Chapter

2B

ackground

Table 2.1.1: Literature overview of mathematical simulation studies on dynamic bone tissue engineering cultures in bioreactors.Bioreactor Scaffold properties Scaffold material Simulation / method Experimentally

validatedReference

Rotatingwall vessel

Microcarrier beads, diameter 175μm Unknown Numerical model No [18]

Hollow microspheres, diameter 100–200μm Ceramic Recording trajectories of scaffolds Yes [19]

Microcarriers, diameter 1mm, density1.05g/cm3

Polystyrene Numerical model describingtrajectories, PIV

Yes [20]

Hollow microcarriers, diameter 500–860μm,density 0.6–0.99g/ml

PLGA Cell study, numerical model, PIV Yes [4]

Simulation: solid; experiment: porosity 97% PGA 2-D mathematical model, tissuegrowth

Yes [21]

Spinner flask Simulation: solid/porous media; experiment:spheres, diameter 11μm, density 2.54g/cm3

Glass PIV, CFD No [22]

Simulation: solid; experiment: porosity 97% PGA CFD No [23]

Simulation: solid; experiment: porosity 97% PGA CFD Yes [24]

Compression μCT, porosity 95%, pore size 100–500μm PLA, Ti-stabilized CaP glass FE modelling, mechanoregulatoryalgorithm

No [25]

μCT, porosity 30.2–32.2%, pore size unknown CaP-based bone cement,porous glass

FE modelling No [26]

Regular gyroid or hexagonal shape, porosity55% or 70%, pore size unknown

PLA FE modelling, CFD,mechanoregulatory algorithm

No [27]

μCT, irregular, porous, porosity and pore sizeunknown

CaP-based porous glass FE modelling, CFD,mechanoregulatory algorithm

No [28]

μCT, irregular, porous, porosity and pore sizeunknown

CaP-based bone cement FE modelling, CFD No [29]

μCT, irregular, porosity 95%, pore size80–210μm

PLA-glass composite FE modelling, CFD No [30]

μCT, irregular, porosity 90%, pore size250–350μm

PLA FE modelling Yes [31]

16

2.1C

omputationalapproaches

Table 2.1.1 continuedBioreactor Scaffold properties Scaffold material Simulation / method Experimentally

validatedReference

Perfusion Experiment: porosity 78.8%, pore size300–500μm

PLGA Cylindrical pore model Yes [2]

Irregular, porosity 67.5%, pore size 350μm CaP Cylindrical pore model Yes [32]

Irregular, porosity 70%, pore size100–1000μm

Decellularized, trabecularbone (cow)

Cylindrical pore model Yes [33]

μCT, irregular, CG: porosity 99%, pore size350μm; CaP: porosity 60%, pore size 96μm

CG, CaP Cylindrical pore model, CFD No [34]

Porous continuous medium, porosity 61%,1-D model solid

Titanium alloy 1-D model, porous continuousmedium

No [35]

Modelled, regular, porous scaffold, pore size0.1mm

Unknown Cellular automaton No [36]

Modelled, regular, honeycomb pattern,porosity 59%, 65%, 77%, 89%, pore size50μm, 100μm, 150μm

Polymeric CFD No [37]

Modelled, regular, porosity 60–80%, pore sizeunknown

Polymeric CFD No [38]

Simulation: modelled, solid; experiment:porosity 89%, pore size unknown

Polyethylene CFD Yes [39]

μCT, irregular, porosity and pore sizeunknown

Human trabecular bone CFD No [40]

μCT, irregular, porosity 80–95%, pore size215–402.5μm

PLA CFD No [41]

μCT, irregular, Ti: porosity 77%, pore size280μm; HA: porosity 73%, pore size 270μm

Titanium, hydroxyapatite CFD No [42]

μCT, irregular, porosity 77%, pore size100μm

Polyesterurethane foam CFD No [43]

17


Spinner flask bioreactors

Like the rotating wall vessel bioreactor, the spinner flask bioreactor uses convectionto ensure that the culture medium surrounding the scaffold is well mixed [2]. Spinnerflask systems consist of a dual-side arm cylindrical flask with a rubber stopper serv-ing as a cover. 3-D scaffolds are attached to needles that pierce the rubber stopper,fixing them in place within the stirring medium. The distance from the scaffolds tothe stir bar can be controlled by the position of the scaffolds on the needles. Scaf-folds are completely covered with culture medium and a magnetic stirrer is placedat the bottom of the flask (Fig. 2.1.1B), stirring the culture medium [2, 17, 44]. Aspecialized spinner flask design is the wavy-walled bioreactor. This modified spin-ner flask bioreactor is fabricated by altering the radius of conventional spinner flasksby introducing grooves in the wall (Fig. 2.1.1B). This design enhances the mixingof the culture medium while minimizing fluid shear forces [23, 45]. Spinner flaskcultures showed improved osteogenic differentiation compared to static or rotatingwall vessel bioreactor cultures and increased calcium deposition at the scaffold’ssurface [2, 17, 46]. Despite these advantages, spinner flask cultures showed sparsedistribution of cells and mineralized ECM was only located at the outer surfaces ofthe 3-D scaffolds [17,46,47].

Compression bioreactors

Compression bioreactors are intended to mimic the macroscopic mechanical stimu-lus of bone in-vivo (Fig. 2.1.2). They are made up of a compression chamber with apiston, which applies compressive loads directly to the scaffold (Fig. 2.1.1C) [48,49].Compression loading closely represents the in-vivo mechanical stimulation of bonecells [30]. Compression studies have been shown to improve cell ingrowth, ECMsynthesis and alkaline phosphatase (ALP) activity [5, 24, 50]. The upregulation ofALP in-vitro has been generally associated with the onset of osteogenic differenti-ation [51]. However, there are still large differences between different papers aboutthe ideal level, duration and frequency of the applied loading condition.

Perfusion bioreactors

The perfusion bioreactor aims to mimic the microscopic mechanical loading of bonein-vivo (Fig. 2.1.2) [52]. Perfusion bioreactor systems pump culture medium throughthe scaffold’s interconnected pores and the scaffold is press-fitted into a culturechamber (Fig. 2.1.1D) [53–55]. Flow perfusion enhances the mass transfer at theinterior of the 3-D scaffold and exerts shear forces on the cultured cells. Several

18


Fig. 2.1.1: The four most prevalent bioreactor principles used in bone tissue engineering. (A)Rotating wall vessel bioreactor: two concentric cylinders rotate at two different velocities (v1 andv2) to accelerate the culture medium between the two cylinders. (B) Spinner flask bioreactor:the culture medium is rotated with a magnetic stir bar. 1, schematic top view of conventionalspinner flask; 2, schematic top view of wavy walled bioreactor. (C) Compression bioreactor: apiston applies direct compression load on the scaffold construct. (D) Perfusion bioreactor: 1, theperfusion system consists of a pump, a media reservoir and the bioreactor housing the scaffold; 2,the scaffold is press-fitted into a perfusion chamber to ensure medium flow through the scaffold.

19


studies have shown that the increased mass transport led to improved distributionof ECM throughout the 3-D scaffold, increased cell number, enhanced expression ofthe osteogenic phenotype and improved mineralized ECM deposition, compared toother bioreactor systems [6, 17,53,54,56].

Although dynamic bioreactors could overcome diffusion limitations at the surfaceof a scaffold, only the perfusion bioreactor was able to eliminate diffusion limitationsinside a scaffold [46, 53]. Consequently, the perfusion bioreactor seems to be a veryuseful dynamic culture technique for bone tissue engineering [2] and is the mostcommonly used dynamic bioreactor nowadays [57].

Fig. 2.1.2: Schematic overview of macro-and microscopic loading of bone in-vivo. A macroscopiccompression load acting on the bone leads to compression and tension regions in the lacuno-canalicular network. The interstitial fluid in the lacuno-canalicular network flows from compressionregions to tension regions. This leads to shear stresses acting on the osteocytes sitting in the lacunasof the lacuno-canalicular system.

2.1.3 Simulation techniques

Simulations are widely performed in bone tissue engineering today [29, 42, 43, 58].Generally, simulations in bone tissue engineering can be divided into simulations ofthe mechanical environment [19,22,30], simulations of the biological environment [36]or simulations of both environments [21, 25,31].

20


The mechanical environment is the sum of all physical forces acting in a system.Early simulation strategies of mechanical environments in bioreactors were basedon simple numerical models. These models solve a problem and give a solution innumbers. Numerical models were most prevalently used to simulate the motion ofscaffolds in rotating wall vessel bioreactors [4, 18, 20]. The advantages of numericalmodels are that they are easy to solve, can be calculated in a short time and give arough estimate of the solution. The latter, however, is also a major disadvantage ofnumerical models. Usually, numerical models apply a lot of rough assumptions, e.g.rigorous geometrical simplifications. This leads to inaccurate results that are very farfrom the reality. Experimental approaches analysing the motion of a fluid or scaffoldsin a bioreactor evolved in the late 1990s and were used until the early twenty-firstcentury. Particle image velocimetry (PIV) is one of these approaches. PIV is anoptical method to determine the velocity field of moving fluids. The speed of scaffoldscan be calculated from the temporal description of the scaffold motion tracked byPIV. Compared to numerical models, experimental approaches are measurement-based and resemble reality more closely. Major limitations arise due to measurementdevices. In the case of PIV, the image sensor limits the resolution and when usingone sensor the method is limited to two dimensions (2-D) [20].

The evolving bioreactors, spinner flask, compression and perfusion, hold the scaf-folds fixed in place and the motion of the scaffolds is no longer an issue. In 2001 asimple mechanical model, the cylindrical pore model, was introduced for the estima-tion of shear stresses in perfused porous structures [2]. It is based on the assumptionsthat flow is uniformly distributed across a structure surface of a given diameter andthe flow of the culture medium is parabolic, and the pores are represented as a bun-dle of parallel cylindrical pores with a diameter equal to the average pore diameter.Wall shear stresses can then be calculated according to:

τwall = 8μVm/d (2.1)

where μ is the viscosity of the fluid, Vm is the mean velocity of the fluid in thepores and d is the mean diameter of the pores. The cylindrical pore model is stillfrequently used today [33, 34]. It gives researchers a quick indication of approxi-mate values of shear stresses in porous scaffolds under perfusion. The geometricalsimplifications, however, do not resemble the mostly complicated, porous and inter-connected geometry of 3-D scaffolds. Additionally, shear stresses in highly irregularscaffolds must be described by a shear stress distribution, as shear stresses in smallpores are lower than in large pores [37].

Over time, mechanical models became more complex. FE modelling is a numerical

21


technique for finding the solutions to differential and integral equations. For FEmodelling, a geometrical shape is subdivided into a finite number of small elements.The displacement of the elements under loading is then modelled for each elementand stresses and strains are computed. A subgroup of FE modelling is computationalfluid dynamics (CFD). In CFD a fluid flow through the void space of a shape issimulated, and fluid shear stresses are then calculated from the fluid velocities. Forall FE models the underlying geometries are obtained either directly by imaging orwith the help of computer-aided design (Fig. 2.1.3). FE models have the advantagethat they are much more accurate than simple analytical and numerical models. Theaccuracy of the geometry is highly increased because it is based on a real geometry.FE modelling is a powerful tool and is able to predict local mechanical effects, evenat cellular level [59]. Nevertheless, calculating mechanical environments throughcomplete scaffolds is far from evident because of the lack of computational power,especially in CFD models [42]. Today, models often no longer reflect a whole systembecause they have been reduced in size or accuracy.

In addition to the mechanical environment, the biological environment in biore-actors can be simulated. The biological environment includes concentrations ofnutrients and waste products, tissue growth and cells. Simulations of the biologicalenvironment in bioreactors started in the late 1990s with numerical simulations ofbone growth, based on mechanobiological models [21, 33, 36, 60]. A mechanoregula-tory algorithm was first introduced by [61]. Briefly, mechanical stimuli are calcu-lated with FE modelling in each element of a material. The tissue phenotype is thendetermined according to predetermined tissue thresholds. Material properties areupdated according to the determined tissue phenotype, because each tissue pheno-type has different material properties. The differentiation thresholds can be adaptedto match other tissue phenotypes or even to under-or overloading [25]. Originally,the mechanoregulation theory was strain-based, but was further adapted to takeshear stresses into account [27, 28]. The downside of the mechanoregulation the-ory is the choice of threshold levels and material properties. The growing tissue ishighly irregular in shape and composition. Additionally, the stiffness of one tissuetype changes depending on its phenotype (e.g. cortical/cancellous bone). Hence,it is not valid to set only one threshold for an entire structure. Despite these lim-itations, very good results have been observed for simulations of osteogenesis inscaffolds using the mechanoregulatory model [30]. Several other models have beenadapted and individualized to simulate specific attributes of bone cell cultures inbioreactors. An organic crystal growth model was used to model tissue growth overtime [21], mass conservation, diffusion-convection, and enzymatic kinetics were used

22


Fig. 2.1.3: Meshed scaffold structures. (A) Silk fibroin scaffold: the scaffold structure was obtainedby microcomputed tomography imaging at a resolution of 6μm; the scaffold was coarsened for themeshing procedure. Reprinted with permission from Simpleware Ltd (Exeter, UK). (B) Gyroidshape and (C) hexagonal prism: these two structures were built artificially and meshed with atriangle surface mesh. Reprinted from Olivares et al. [27], with permission from Elsevier.

to model oxygen concentration as well as oxygen transport in scaffolds [35, 62] anda random walk algorithm simulating cell migration [36], just to name a few. Theapplication of these mathematical models will now be described for the four majortypes of bioreactors used in bone tissue engineering, with respect to mechanical, bi-ological or combined mechanical and biological environments. For some bioreactortypes, literature is lacking for one or more of those environments. In consequence,the three environments are not discussed for all bioreactor types.

Simulation of rotating wall vessel bioreactors - mechanical environment

Boyd [18] was one of the first to model the flow field of circulating culture mediumin a rotating wall vessel bioreactor. He showed that the circulation in the middleof the rotating wall vessel was poor and that maximum shear stresses occurring atthe scaffold’s surface were in the range 0.0002–0.0013Pa. Nevertheless, the study

23


was purely mathematical and did not contain comparisons to, or implementation of,experimental observations. About 10 years later Qiu et al. [19] calculated maximumshear stresses on scaffolds, based on experimentally recorded trajectories (Fig. 2.1.4).A camera recorded the locations of the scaffolds while the rotating wall vessel wasturning. Based on trajectory recording, the maximum shear stresses were calcu-lated to be around 0.06Pa, which were about 50 times higher than the maximumvalues observed by Boyd. A bone cell line culture did not show any detrimentaleffects to the cells, and cells were able to attach to the scaffolds and form ECMand mineral nodules [19]. A similar study developed a numerical model to describethe trajectory of a scaffold in a rotating wall vessel bioreactor, which was validatedwith PIV; the authors could show that the predicted results from the numericalmodel were in excellent agreement with experimental measurements [20]. The re-sults were compared to an experimental cell study performed under similar cultureconditions. A bone cell line was cultured for 7 days in a rotating wall vessel bioreac-tor on poly(d,l-lactic-co-glycolide acid) (PLGA) scaffolds. It was shown that cellspenetrated as deep as 800μm into the scaffold [4]. This is four times higher thana penetration depth of 200μm under static conditions in a scaffold with a similarpore size distribution [63]. ALP activity was higher for cells cultured in the rotatingwall vessel bioreactor than in static culture after 7 days. This confirmed the resultspredicted by the simulation: (a) the motion of the scaffold in the bioreactor createda convective flow at the scaffold’s surface, which could have led to the increasedpenetration of the cells; and (b) maximum shear stresses were determined between0.27Pa to 0.47Pa, which could have led to the increased ALP expression [4]. Thecombination of simulations with experimental observations in rotating wall vesselbioreactors built the basis for further simulations on tissue-engineering cultures inother bioreactors. The simulations of rotating wall vessel bioreactors were able to re-veal distinct influences on bone cell cultures; however, the results are of a qualitativenature and indicate only a rough estimation of culture properties.

24


Fig. 2.1.4: Recorded trajectories of a scaffold in a rotating wall vessel bioreactor (inertial frame).The scaffold moved in a spiral fashion towards the centre of the bioreactor; consecutive pictureswere taken every 10min. Reprinted from Qiu et al. [19], with permission from Elsevier.

Simulation of rotating wall vessel bioreactors - biological and mechanicalenvironment

Lappa [21] introduced a 2-D mathematical model in 2003 including not only mechan-ical properties but also bioreactor-specific features for growth, nutrient transport,nutrient transformation into organic tissue, mass variation of the specimen, andgrowing tissue. A combined model simulated mass transfer at the scaffold surfaceto determine the concentration field of glucose in the liquid phase. The model wasfurther adapted by including biological tissue growth rate to take the three main as-pects of growth behaviour into account: (a) availability of nutrients; (b) slow surfacekinetics; and (c) the effect of surface shear stress. The simulations were compared toan experimental study, using the same bioreactor [64]. It was shown that especiallythe corners and edges of the scaffold were supplied with a higher amount of glucose.The growth simulation was able to reproduce the morphological evolution observedby Obradovic et al. [64] (Fig. 2.1.5). Fluid shear stress distribution was the majorfactor influencing oxygen absorption kinetics and tissue growth. Shear stresses atthe scaffold’s surface were non-uniformly distributed and changed over time due totissue growth. In regions with lower shear stresses, tissue growth was increased,likely because fluid close to stagnation allows for better surface absorption kineticsof glucose. However, these results are in disagreement with other studies showingincreased growth with increased shear stresses. This study was one of the first con-cerning the biological environment and the influence of growing tissue as a moving

25


boundary condition in rotating wall vessel bioreactors. The results, however, havethus far only been observed in 2-D cartilage constructs [21].

Fig. 2.1.5: Figure 5. Comparison of numerical and experimental results of tissue growth. (A)Simulation of tissue growth pattern and progression. Reprinted from Lappa [21], with permissionfrom Wiley-Blackwell. (B) Histological cross-section of cartilage construct cultured for 6 weeks.The morphology of the simulated construct is in very good agreement with the experimental resultsof Obradovic et al. [64], shown by six characteristic points (A-F) along the construct surface.Reprinted from Obradovic et al. [64], with permission from Wiley-Blackwell.

Simulation of spinner flask bioreactors - mechanical environment

Sucosky and colleagues [22] introduced one of the first simulations describing theflow field in a spinner flask bioreactor. A CFD approach and experimental PIVwere compared to assess the validity of the results. Both the simulation and theexperiment were conducted under the same culture conditions. CFD simulationsshowed that the velocity components differed approximately 20% from the resultsof the PIV, but both the trends and amplitudes of the velocities were similar. Thesimulated maximum shear stress of the CFD was 0.21Pa and was located near thelower surface of the scaffold as well as along the vertical wall. Computational sim-ulation of the shear stresses was in agreement with the experimental results. Themaximum shear stress of the PIV was 0.25Pa and occurred at the lower surface ofthe scaffold, which was expected for a spinner flask with a magnetic stir bar placedat the bottom. The maximum shear stress level differed by about 16% from thesimulation, because the vertical position of the scaffold in the experimental modelwas lower, and therefore the construct experienced higher fluid velocities because ofthe vicinity to the stir bar. This study showed that differences between experimental

26


and computational simulations can be up to 20%. The outcomes were very encour-aging, but they only constituted an initial step in the design of a reliable tool toinvestigate effects of culture medium convection [22] and independent experimentalverification is missing.

Bilgen et al. [23] and Bueno et al. [45] performed simulations using a wavy-walledbioreactor. They characterized the effect of the wavy walls on flow patterns byCFD and compared the results to the simulation of a regular spinner flask. Thewavy-walled bioreactor showed a more uniform distribution of flow patterns in thebioreactor. The velocity in the middle, where the scaffolds were placed, was dou-bled compared to a regular spinner flask, indicating a beneficial effect of the wavywalls on the mixing of culture medium. Compared to the regular spinner flask, theresulting average shear stresses increased by 6%. The authors predicted the largershear stresses to enhance aggregation of cartilaginous cells and increase nutrient andgas transport at the scaffold’s surface. Bueno et al. [45] verified the CFD resultsby performing a cell study using calf chondrocytes. After 4 weeks of culture, chon-drocyte proliferation and matrix deposition were enhanced, confirming the positiveeffects of the wavy-walled design. This study revealed how important the designof the bioreactor itself is. Not only should scaffold geometry be incorporated insimulations, but also the geometry of the bioreactor itself.

In general, spinner flask bioreactors were used primarily for cartilage tissue engi-neering. Initially, the mechanical environment in spinner flask bioreactors seemed tobe beneficial for bone tissue engineering cultures as well, although compression andperfusion bioreactors have evolved, leading to better results in bone tissue culturesthan with spinner flask bioreactors.

Simulation of compression bioreactors

Compression bioreactors are often combined with the application of perfusion [29,31]. This combination closely resembles the in-vivo loading of bone: macroscopiccompression load and microscopic perfusion flow (Fig. 2.1.2). Therefore, combinedcompression-perfusion studies will be mentioned in the following sections, along withconventional compression studies.

Simulation of compression bioreactors - mechanical environment

Lacroix et al. [26] investigated the effect of the load transfer from three differ-ently prepared calcium phosphate (CaP)-based scaffolds to the cells. Scaffolds werescanned using μCT and scans were divided into smaller cylindrical volumes of inter-est (VOIs), due to computational limitations. A compressive axial strain of 0.5% was

27


applied on the upper part of the scaffolds. The obtained results showed a variation instrain values by a factor of 3–4 between the different materials, but the preparationof the CaP-based scaffolds did not influence subsequent mechanical behaviour.

Sandino et al. [29] showed that compressive loads applied to irregular scaffoldsled to different strain levels throughout the scaffold, according to its morphologyand geometry. Under 0.5% global strain, local compressive strains were in the range0.2–0.6%. High changes in fluid velocity were observed at 1, 10 and 100μm/s, withregions of almost no flow and regions with high-velocity fluid flow. The majority ofshear stress values were around 5∗10−7Pa. Maximum shear stress values were about800 times higher than mean stress values (0.0004Pa). The distribution of stresses andstrains was highly heterogeneous thoughout the scaffold structure [29,31]. Irregularscaffolds are mostly used in bone tissue engineering today, due to the variety inmanufacturing methods and because they mimic the geometry of cancellous bone.However, this irregularity leads to a high variation of the mechanical loads actingon cells in-vitro, which makes it hard to control the mechanical stimulation of cells.Such studies should be combined in the future with in-vitro studies to improve theunderstanding of tissue differentiation in a scaffold.

Milan et al. [30] analysed the mechanical environment induced by dynamic com-pression loading and perfusion flow on the basis of a μCT scan of a polylactic acid(PLA)-glass composite scaffold. A steady fluid flow of 100μm/s and a dynamiccompression of 5% at a strain rate of 1/s were applied to the PLA-glass scaffoldwithin a cylindrical bioreactor. The highest fluid-flow velocities were found in thecentres of the scaffold pores, whereas the lowest velocities were allocated near thepore walls. Mean stress occurring in the glass part was four times higher than themean stress calculated for the PLA part, because the glass part was 20 times morestiff than the PLA part. Large standard deviations showed a heterogeneous distri-bution of stresses and strains within the scaffold. The authors could show that thearchitecture of the scaffolds could have a big influence on mechanical stimulationof seeded cells, but they did not perform a cell study to confirm this statement.The mechanical environment analysis performed is of great interest for bone tissueengineering, because it closely resembles the in-vivo macroscopic and microscopicmechanical stimulation of bone. However, the authors did not simulate both loadingconditions simultaneously. It would be interesting to analyse the combined effects ofsimultaneous compression and perfusion loading. Apart from that, a heterogeneous,composite scaffold design could introduce interesting effects on cultured cells, suchas increased bone density at stiffer sites, or decreased bone density at softer sites.

28


Simulation of compression bioreactors - biological and mechanicalenvironment

Milan et al. [25] simulated cell differentiation under dynamic compression. The ge-ometry of a PLA-glass scaffold was reconstructed from μCT scans and a biologicalmaterial containing fluid, cells and matrix was simulated to fill the scaffold pores.A mechanoregulation algorithm was applied to determine the differentiation of thebiological material. The algorithm predicted formation of immature and maturebone for compressive strains of 0.5–1% at strain rates of 0.0025–0.005/s in the mid-dle of the pores. Cartilage and fibrous tissue formation was predicted at higherstrain levels and close to the pore walls [25]. This simulation study was a fur-ther development of their study in 2009, where the authors simulated compressionand perfusion loading separately. The combination of compression and perfusionloading closely resembles the in-vivo loading conditions (Fig. 2.1.2). Concerningtissue-engineering applications, the simulation of biological tissue completely fillingthe scaffold pores is not applicable for early time points, because cells and matrixwould not yet be filling the pores completely. Therefore, this model should only beused for simulating late stages of tissue-engineering applications where the pore vol-ume is already completely filled with tissue. The results of Milan et al. [25] were inagreement with the study of Olivares et al. [27]. Olivares and colleagues [27] studiedthe interactions between scaffold morphology and applied culture conditions on tworegular scaffold structures: gyroid and hexagonal (Fig. 2.1.3B, C). The resultingstrains and fluid shear stresses were calculated at an axial strain of 5% and an inletvelocity of 1mm/s, respectively. The mechanoregulation theory of Prendergast etal. [61] was adapted to take the combined effects of strains and shear stresses intoaccount. Strain values of 0.5–2.5% showed a prevalence of osteogenic differentiation,whereas chondrogenic differentiation appeared at 2.5% strain. An inlet velocity of0.001mm/s was favourable for bone stimulation in both geometries. Hexagonal scaf-folds showed fewer changes in fluid path, leading to a better distribution of strainsfor bone phenotype. Beside the fact that artificial scaffolds do not represent phys-iological bone geometry, they are often advantageous to simulating the effects ofmechanical stimulation in a controlled manner. The study presented is one of thefirst combining simultaneous compression and perfusion stimulation. However, theresults were again presented concerning only compression or perfusion effects. Un-der physiological conditions, a bone is always affected by both loading regimes andthis should therefore be considered in the future.

Sandino and Lacroix [28] confirmed the results of their earlier study from 2008, inwhich they simulated the mechanical environment of compression bioreactors. Ad-

29


ditionally, they determined tissue differentiation under 0.5% strain load and 10μm/sperfusion flow, using the mechanoregulation theory. Stimulus distribution was highlyheterogeneous. Some stimuli increased up to the load range of cell death in regionswith very high shear stresses. This effect occurred because the predicted tissue for-mation decreased the porosity until the pores were completely filled with tissue [28].This effect can also be observed in-vitro. Growing tissue within a scaffold leads tosmaller or obstructed pores, which results in increased shear stresses acting on thecells in these pores, assuming the cells to be sitting on top of the growing ECM. Theglobal mechanical stimulation should therefore be adapted to the growing tissue.This study demonstrated one possible solution to simulating in-vitro tissue growthin porous scaffolds and, according to these results, the in-vitro culture conditionscould be adapted.

Baas et al. [31] combined an FE model with an experimental bone cell study [48].The same experimental conditions for macroscopic compression were simulated aspreviously described [48]. Scaffolds were scanned with μCT before the start of thecell study. The seeded scaffolds were maintained in static culture for 2–3 weeksand were then dynamically loaded in a compression-perfusion bioreactor at 1.5%strain and 1Hz for 1h daily, for 1 week. In addition to the compression loading, thescaffolds were perfused continuously at a rate of 0.1ml/min. After a total of 4 weeksof culture time, all the scaffolds were scanned again. Both datasets from the pre-and post-scans were then compared for each scaffold separately to correlate localprincipal strain at the start of the culture with local mineralization at the end ofthe culture (Fig. 2.1.6). The average value of principal strain before the culture wassignificantly higher at sites where mineralized ECM had formed compared to siteswhere no mineralized ECM had formed. The results showed that bone cells in a 3-Denvironment are sensitive to surface strain, leading to mineralized ECM formationin locations with higher local strains. This study is one of the few studies showing adirect connection between local mechanical stimuli and mineralized tissue formationin a 3-D environment [31].

An important future step in simulating the behaviour of compression bioreac-tors is the combination with perfusion loading, because until now compression andperfusion have been modelled separately [27–29].

Simulation of perfusion bioreactors - mechanical environment

Goldstein et al. [2] predicted shear stresses with the cylindrical pore model for aporous PLGA scaffold. The applied flow of 0.03ml/s/scaffold led to a shear stressof 0.034Pa. Experimental results showed that this fluid flow applied to the cultured

30


Fig. 2.1.6: Micro-computed tomography images of a scaffold before and after culture: (A) emptyscaffold before culture; (B) strain distribution in the empty scaffold; (C) scaffold after culturewith mineralized nodules formed (orange). Mineralized nodules have formed at locations of higherstrains. Reprinted from Baas et al. [31], with permission from Elsevier.

cells improved cell distribution in the scaffold and led to increased osteogenic dif-ferentiation [2]. Similar studies were performed by others [32, 33]. Vance et al. [32]exposed bone cells seeded on CaP scaffolds to high-rate oscillatory flow, low-rate per-fusion flow and static culture conditions. With the use of the cylindrical pore model,the shear stresses in continuous perfusion at a flow rate of 0.025ml/min and underoscillatory flow at 1Hz, with a peak flow rate of 40ml/min, resulted in 0.0007Paand 1.2Pa, respectively. At both shear-stress levels, the release of prostaglandinE2, which is thought to have an anabolic effect on bone but is also an inflamma-tory marker, was significantly increased. DNA content was not affected, despitethe high fluid flow velocity [32]. Grayson et al. [33] investigated the influence ofperfusion flow velocities in the range 80–1800μm/s on human mesenchymal stemcells (hMSCs) seeded on bone scaffolds, calculating shear stresses using the cylin-drical pore model. Shear stresses increased with increasing fluid flow velocity from0.0006Pa to 0.02Pa. An optimal range of flow velocities resulting in the highest ECMdeposition for hMSCs seeded on bone scaffolds was determined to be between 400–800μm/s. The range of shear stresses shown to be beneficial for bone cell culturesis very broad, with the highest value being 2000 times higher than the lowest value(0.0006Pa–1.2Pa). This shows the low specificity of the cylindrical pore model, dueto numerous assumptions. Jungreuthmayer et al. [34] compared the results of thecylindrical pore model to a μCT-based CFD model. Analytical results of the wallshear stress using a velocity of 235μm/s showed values of 0.022Pa and 0.903Pa forcollagen-glycosaminoglycan (CG) and CaP scaffolds, respectively. Mean wall shearstresses acting on CG scaffolds were 0.019Pa, and CaP scaffolds experienced a mean

31


wall shear stress of 0.745Pa, as determined by CFD. These results suggested thatthe analytical model of Goldstein et al. [2] overestimates the wall shear stresses,especially at higher fluid flow velocities. This could lead to a suboptimal stimula-tion of cells when using the analytical model to determine culture parameters. Athigh flow velocities the difference between the two models is >20%, confirming theassumption of the low specificity of the cylindrical pore model.

Scaffold properties such as connectivity, porosity and pore size play an importantrole in perfusion cultures. Simulations of regularly shaped scaffolds are a straight-forward method to investigate different scaffold properties. Boschetti et al. [37]simulated the influence of porosity and pore size on mechanical environment with asimple 2-D scaffold model. The scaffold was modelled by subtracting a solid spherefrom a concentric solid cube. The dimensions of the sphere and the solid cube werevaried to obtain different pore sizes and porosities. The velocity map looked thesame when simulating with constant porosity, independent of pore size. The localvelocity gradient was bigger for smaller pores. As expected, the wall shear stresseswere higher with a smaller pore size. Wall shear stresses appeared to be indepen-dent of the porosity at constant pore size, except for a small area around the inletand outlet, where high shear stresses were observed (Fig. 2.1.7). The values of shearstresses were roughly constant with increasing porosities but increased with decreas-ing pore size, which shows that the pore size is a parameter strongly influencing thepredicted wall shear stress. This study had the advantage that the observed resultswere qualitative and could be easily transferred into 3-D and more complex scaf-folds, where the same rules apply, as Yao et al. [38] showed; they modelled an entirescaffold and compared the results against those of Boschetti and colleagues [37], whomodelled only a microdomain of the scaffold containing 27 pores. Yao et al. [38] con-firmed that the velocity map showed the same trend with constant porosity: shearstress decreased with increasing porosity and marginal regions showed higher shearstresses than the rest of the scaffold.

CFD simulations are widely used in combination with μCT [40–43]. CFD modelsare especially suitable for predicting fluid velocities, fluid pressure and the fluid shearstresses acting on cells [27,37,38]. Porter et al. [40] performed one of the first studiescombining a μCT scan of human trabecular bone, defining the physical boundaryconditions for the CFD model. The highest flow velocities were observed at thecentre of small orifices and the lowest flow velocities were observed at the scaffoldsurface and bioreactor chamber walls. These results confirmed the basic conceptualsimulations of Boschetti et al. [37] on a physiological sample. Local shear stressesexperienced by cells at a constant flow rate can be vastly different, ranging from 0Pa

32


Fig. 2.1.7: Shear stress maps on the surface of a pore (a quarter of a pore surface is shown). Shearstress values increased with decreasing pore size and porosity, and reached the highest values atthe pore inlet. Reprinted from Boschetti et al. [37], with permission from Elsevier.

to 0.0002Pa, because of the irregular scaffold geometry. A similar study confirmedthese results and also investigated the influence of scaffold manufacturing techniqueon shear stress distribution. As shown previously, decreasing pore size and porosityled to a more constricted flow field and increased shear stresses. Defective scaffoldarchitectures led to increased fluid flow through the area where solid material wasmissing, because fluids choose the path of least resistance. Therefore, shear stressesincreased near the defect site [41].

An important consideration for CFD simulations is the definition of boundary con-ditions. Maes et al. [42] investigated the effect of different sizes of VOIs. Wall shearstresses were about 30% higher in small VOIs (cubic, 1mm side length) compared tolarger VOIs (cubic, 1.5mm side length). A possible explanation for this phenomenonwas that with a smaller VOI scaffold heterogeneity was captured poorly, and thatthese models experienced a larger influence from boundary conditions. The authorsconcluded that the minimal model size should be at least twice the pore size, butas large as possible to cover the heterogeneity and the actual micro-architecture ofthe scaffold completely. Compared to other simulations in the context of porousmaterial properties, the minimum size of a representative VOI for a continuum do-main was determined to be at least 3.5 times the pore diameter [65, 66]. Defining

33


the smallest valid VOI representing the whole scaffold is of significant importancefor reducing the computational power of very large models.

Unfortunately, experimental data to validate perfusion bioreactors is mostly miss-ing. One paper quantitatively evaluating the effects of shear stresses on hMSCs,both mathematically and experimentally, is the study of Zhao et al. [39]. At a flowrate of 1.5ml/min the velocity in the scaffold decreased from 2.0μm/s at the sur-face to 0.1μm/s at a depth of 70μm. The corresponding shear stresses decreasedapproximately one order of magnitude. For a fluid flow velocity of 0.1ml/min, thevelocity and shear stresses at the surface were all lower than 0.1μm/s and 10–5Pa,respectively. A cell study was performed with identical settings. Poly(ethyleneterephthalate) scaffolds were seeded with hMSCs and were cultured at a flow rateof either 0.1ml/min or 1.5ml/min for 20 days. The results of the cell study showedthat perfusion flow, even at very low flow velocities, had a significant effect on thecultured hMSCs. Low flow velocity led to higher cell numbers, whereas higher flowled to statistically increased ALP activity and calcium deposition [39]. These resultswere in agreement with previous studies showing the same effects of fluid flow inosteogenic cultures [12,67,68].

Simulation of perfusion bioreactors - biological environment

Truscello et al. [35] and Pierre and Oddou [62] investigated the effect of perfusionflow on oxygen distribution and transport within perfusion bioreactors. Truscelloet al. [35] simulated oxygen distribution in a perfusion bioreactor and determinedthe critical length of the scaffold to guarantee a given target range of oxygen ten-sion. Cells were simulated either as an attached cell layer on the scaffold wallsor suspended in the culture medium. The critical scaffold length was determinedto be proportional to the inlet velocity and inversely proportional to the cellularconsumption rate and cell density, using a 1-D model [35]. Pierre and Oddou [62]investigated oxygen concentration subject to the inlet velocity for a large bone im-plant (d=10mm, l=25mm). Cells were modelled as a monolayer attached to thescaffold surface. The simulation showed that having a local oxygen concentrationsufficient for cell metabolism, 64% of the scaffold surface was considered to be loadedunder detrimental mechanical conditions (shear stresses >10−3Pa). Considering alower inlet velocity, scaffold surface is loaded under adequate mechanical conditionsbut oxygen concentration on 32% of the scaffold surface is under hypoxic conditions.Compared to the study of Truscello and colleagues [35], this study shows the impor-tance of not looking only at one factor (e.g. mechanical loads) of a cell culture study.The study presented is very promising for a transformation into a 3-D model, because

34


it simulates the multiple biological properties of a cell culture: (a) cells; (b) oxy-gen concentration; and (c) oxygen consumption [62]. Galbusera and colleagues [36]performed an even more detailed study, including cell population dynamics. Theydetermined oxygen transport and consumption, hydrodynamic environment and cellmovements for the prediction of in-vitro tissue growth. All culture conditions butthe hydrodynamic environment were simulated by a cellular automaton. A cellularautomaton usually consists of a regular grid of cells. Briefly, the scaffold was dividedinto equally spaced points defined as scaffold or fluid. Each of these points could hostone cell and was able to be in one of three states: (a) moving at a migration speedof 1μm/s; (b) stationary, due to a collision with another cell; and (c) stationary dueto a collision with the scaffold wall. If the cell is in state (a), it keeps moving. Celldivisions were also taken into account (Fig. 2.1.8). A homogeneous oxygen concen-tration of 0.2nM/ml was initialized in the whole fluid. The simulations showed afirst stage of exponential cell growth, followed by a deceleration when the volumefraction of occupied nodes was >0.2 and a final decrease of growth rate due to thefilling of available spaces. A higher cell number had a significant effect on oxygenconcentration, as cell consumption led to decreased oxygen concentration. Perfu-sion flow reduced the drop of oxygen concentration between the inlet and outlet, ascompared to statically cultured scaffolds. The study focused on oxygen concentra-tion only while nutrient diffusion and waste removal were not taken into account.Another important limitation was the absence of an ECM in the simulation, whichmodifies the cellular microenvironment, especially the oxygen concentration. Addi-tionally, the influence on fluid velocity due to filling of the available pore space wasnot investigated [36].

The increasing number of papers published on bone tissue engineering with perfu-sion bioreactors reflects the importance of this bioreactor type. Experimental studieswere able to show promising effects of perfusion on cell number [12, 39] or mineral-ized ECM formation [6,69]. Simulations strongly support the need to determine thebest influencing parameters for bone tissue cultures.

2.1.4 Discussion and future directions

The application of simulations in dynamic bone tissue engineering using bioreactorshas evolved over the last two decades as a crucial research field, building bridgesbetween biology and engineering. Since the early 1990s simulations have played animportant role in bone tissue engineering, advancing from simple numerical mod-els [2,18] to very complex, but more accurate, computational simulations [28,31,41].Most of the studies presented in this review are still in their infancy concerning

35


Fig. 2.1.8: Representative image showing cell migration and proliferation in a two-dimensional do-main. Cell division, collision between cells and collision between a cell and the wall are represented.Reprinted from Galbusera et al. [36], with permission from Taylor&Francis Group.

their expected use in predicting the development of bone-like tissue cultures overtime. The future of simulations of bone tissue engineering in bioreactors is promis-ing. Technological improvements such as imaging techniques with higher resolutionand increased computational power will enable simulations of whole scaffolds andan increased number of parameters for more accurate and physiologically relevantsimulations in the future.

The understanding of tissue construct development and growth must be enhancedto catch the relationship between the bioreactor’s environment and cellular re-sponses. The environment provided by bioreactors needs to be precisely controlledto produce: (a) more reproducible methods; (b) more realistic in-vitro studies, im-itating environmental cues acting on cells in-vivo; and (c) precise and controlledapplication of environmental cues to improve and optimize cell response [57].

Most simulations modelling biomechanical laws remain poorly developed [16].This is mainly attributable to moving boundary conditions and the assumptionsmade in computational models. Simulations performed in 2-D do not reflect thegeometry of bone in-vivo. It was shown in-vitro that the dimension has a majorinfluence on the effect of mechanical stimulation [70]. Despite the fact that the true

36


geometry of the scaffolds can be obtained by various imaging techniques, regularlyshaped scaffolds were often used to perform simplified simulations [23,27,37]. Thesescaffolds do not resemble the architecture present in native bone or the scaffoldsused in current in-vitro experiments.

Cells were mostly neglected in the models [28, 30] or, if cells were modelled, nocomparisons were made to a no-cell situation [36]. Cells have been modelled asattached to the surface of the scaffold in a regular manner or evenly suspendedin the culture medium. in-vitro it is almost impossible to seed cells uniformly on ascaffold, and assuming evenly suspended cells in simulation models does not resemblereality. It was shown that the distribution of cells does have an important effect onoxygen concentration [35], but the effect of cells, and especially the ECM producedby them, on the mechanical environment has not yet been investigated.

Simulations of bone tissue cultures are often stationary and the formation ofnew ECM and neo-tissue over time is neglected. Growing tissue is a major issue,especially in perfusion cultures. As tissue mass increases, the interconnected poresof the scaffold become more and more obstructed. This process gives rise to anincrease in mechanical shear stress acting on the cells. If the mechanical stimulationis not adapted to the growing tissue, it may lead to detrimental effects on the cells.Growing tissue can also lead to a change in mechanical properties of the wholeconstruct. Especially under compression, growing tissue would have an influence onthe mechanical stimuli a cell experiences. Another effect is the deformation of thescaffold in perfusion studies, an effect that is thought to be small and negligible [30].In turn, hydrodynamic stimulation caused by moving fluid due to compression of thefluid or movement of the load-applying piston is most often neglected in compressionstudies. Nevertheless, if the understanding of a bone tissue culture is to include asmany of the influencing aspects affecting the culture as possible, these parameterswill need detailed investigation in the future. Moving boundary issues remain achallenging task to be solved in future studies.

The implementation of quantitative imaging techniques contributes vastly to com-putational simulations, as only factors that can be determined quantitatively canserve as parameters affecting the simulation. For example, scaffold geometry atthe beginning of an in-vitro culture can be determined and reconstructed as a 3-Dimage; simulations can then be based on these images. Additionally, scaffold param-eters such as porosity, pore size, surface area/unit volume and interconnectivity canbe calculated from reconstructed data. Image-based models have investigated theeffect of scaffold structural properties on bone regeneration [71], their influence onmechanical stimuli distribution [26], fluid flow through the interconnected pores of

37


the scaffold [40] and fluid shear stress within a scaffold [43]. Image-based models ledto more accurate and more realistic models, making the method essential for futurestudies. A particular aim should be to look for a possible strategy to determinethe exact location of cells in a tissue-engineered construct to include actual posi-tion, size and occupied volume of cells in the simulations. An advantage of imagingtechniques is the possibility of obtaining patient- or site-specific data. Prior to im-planting a tissue-engineered construct into a specific anatomical site of a patient, itsmechanical and structural properties can be evaluated by combining computationalsimulations and imaging techniques. Nonetheless, scanning of patients can be dif-ficult at this point, because high-resolution scans (around 10μm) of large volumesare time consuming and the radiation dose for the patient would be too high. Forsimulations investigating processes and scaffold properties at the cellular scale, theresolution must be in the sub-μm range; however, the resolution is still limited byscanner settings, scanning time, radiation dose and a lack of computational powerto solve high-resolution simulations. Image-based models have one critical disad-vantage: large computational power is required to model the already small volumesof irregular scaffolds. Simulating the fluid flow through a complete scaffold at highresolution (around 10μm) with a volume of about 20mm3 is currently far from re-alizable in terms of computational power [42]. In recent studies, a VOI was chosenwhich was considered to be representative of the whole scaffold [29,40–42]. Anothertechnique that was applied to reduce computational power is structure coarsen-ing. Nevertheless, coarsening leads to a loss of information concerning the scaffold’sarchitecture, due to decreased resolution, and was found to be accountable for un-derestimations in wall shear stresses [42]. However, a resolution of 1μm or less isrequired to understand all influencing parameters on a cellular scale, such as scaffoldmicroporosity or surface characteristics. Future work needs to include optimized al-gorithms to reduce computational power, or must be run on supercomputers. Thiswill lead to more complex and bigger computational simulations at higher accuracyand potentially lower cost.

In simulation studies the connection between biology and mechanics must be em-phasized. Most of the studies so far have focused on mechanics only and neglectedthe biology. The effects of nutrient and waste distribution, culture medium con-centration, oxygen concentration and consumption, physical and chemical surfaceproperties and so forth have rarely been included in modelling studies. It is neces-sary to include these effects to effectively model a complete bone tissue-engineeringsystem.

In many cases, experimental data or experimental validation of the simulations is

38


missing. It is very important to show the significance of a simulation by comparisonwith an experimental model. Only experimental models allow the effects of certainparameters on the cellular environment to be verified. More experimental data hasto be produced in the future to validate and test the numerical models.

A general issue in tissue engineering and also in simulating tissue engineeringcultures is the variability between different studies, which makes any comparisondifficult. As influential parameters and boundary conditions are still not determinedand the magnitude of their effect is still unknown, even a small change could havea significant effect in-vitro. Selecting boundary conditions for simulations has ledto additional disparity between the different studies. This complexity should bereduced to make comparisons easier and reduce the number of studies performed.

In the future it will hopefully be possible to determine cellular volume and cellularspread within a scaffold. A prediction of the mechanical load a cell feels and itsresponse in terms of gene expression and ECM production should be possible, anda temporal forecast about the development of the tissue culture will be feasible.The ultimate goal of simulations in tissue engineering is to develop a predictivemodel. If a model can simulate all essential factors acting on a tissue-engineeredsystem, it will be capable of simulating tissue growth and differentiation. Like this,influential parameters can be chosen in advance and can be determined continuouslyand adapted to the actual situation. It may even be possible to automate the wholeprocess, using feedback-controlled mechanisms. If such knowledge can be translatedfrom the in-vitro to the in-vivo situation, even prediction of patient-specific in-vivoperformance might be possible in the future.

2.1.5 Conclusion

The field of simulations of dynamic bone tissue engineering in bioreactors has evolvedrapidly over the last two decades. The simulation approach is multidisciplinary andbuilds a bridge between biology and engineering. However, this bridge is still narrowand a lot of effort must be expended to improve the dialogue between experts inexperiments and experts in simulations. It is crucial to understand the constructdevelopment in response to mechanical loading for the improvement and continu-ation of current tissue-engineering strategies. With the help of simulations, morerealistic in-vitro studies mimicking the in-vivo environment will be possible. Onlywith such studies can we improve our understanding of the biological processes intissue-engineering cultures. Bioreactors play a central role in bone tissue engineer-ing because they provide a high degree of reproducibility, control and a possibilityof automation, and therefore have the capability to improve the quality of engi-

39


neered bone tissue. Simulations will play a major part in advancing the field ofbone tissue engineering by enabling the understanding of causal relations betweenenvironmental cues and the final construct.

Acknowledgements

The authors would like to acknowledge Dr. Davide Ruffoni, who gave scientificinput to this review, and Marie Elise Godla and Samantha Jean Paulsen for theirhelp in checking language and spelling mistakes. We also thank the contributorswho kindly agreed to permit reproduction of the figures. We acknowledge financialsupport from the European Union (EU Project No. FP7-NMP-2010-LARGE-4:BIODESIGN—Rational Bioactive Materials Design for Tissue Regeneration).

References

[1] A. Sittichokechaiwut, J. H. Edwards, A. M. Scutt, G. C. Reilly. Short boutsof mechanical loading are as effective as dexamethasone at inducing matrixproduction by human bone marrow mesenchymal stem cells. Eur Cells Mater,20:45–57, 2010.

[2] A. S. Goldstein, T. M. Juarez, C. D. Helmke, M. C. Gustin, A. G. Mikos. Effectof convection on osteoblastic cell growth and function in biodegradable polymerfoam scaffolds. Biomaterials, 22(11):1279–88, 2001.


[4] E. A. Botchwey, S. R. Pollack, E. M. Levine, C. T. Laurencin. Bone tissueengineering in a rotating bioreactor using a microcarrier matrix system. JBiomed Mater Res, 55(2):242–253, 2001.

[5] V. David et al. Ex vivo bone formation in bovine trabecular bone cultured ina dynamic 3D bioreactor is enhanced by compressive mechanical strain. TissueEng Part A, 14(1):117–126, 2008.

[6] V. I. Sikavitsas, G. N. Bancroft, H. L. Holtorf, J. A. Jansen, A. G. Mikos.Mineralized matrix deposition by marrow stromal osteoblasts in 3D perfusionculture increases with increasing fluid shear forces. P Natl Acad Sci USA,100(25):14683–8, 2003.

40


[7] V. Dixit. Development of a bioartificial liver using isolated hepatocytes. ArtifOrgans, 18(5):371–384, 1994.

[8] L. E. Freed, G. Vunjaknovakovic, R. Langer. Cultivation of cell-polymer carti-lage implants in bioreactors. J Cell Biochem, 51(3):257–264, 1993.

[9] C. R. Halberstadt et al. The in vitro growth of a three-dimensional humandermal replacement using a single-pass perfusion system. Biotechnol Bioeng,43(8):740–6, 1994.

[10] T. Y. Wang and J. H. D. Wu. A continuous perfusion bioreactor for long-termbone-marrow culture. Ann Ny Acad Sci, 665:274–284, 1992.

[11] R. M. Nerem. Tissue engineering in the USA. Med Biol Eng Comput, 30(4):Ce8–Ce12, 1992.


[13] A. J. El Haj and S. H. Cartmell. Bioreactors for bone tissue engineering. ProcInst Mech Eng H, 224(12):1523–32, 2010.

[14] R. P. Schwarz, T. J. Goodwin, D. A. Wolf. Cell culture for three-dimensionalmodeling in rotating-wall vessels: An application of simulated microgravity. JTissue Cult Methods, 14(2):51–7, 1992.

[15] G. Vunjak-Novakovic et al. Bioreactor cultivation conditions modulate thecomposition and mechanical properties of tissue-engineered cartilage. J OrthopRes, 17(1):130–138, 1999.

[16] H. Singh and D. W. Hutmacher. Bioreactor studies and computational fluiddynamics. Adv Biochem Eng Biotechnol, 112:231–249, 2009.

[17] V. I. Sikavitsas, G. N. Bancroft, A. G. Mikos. Formation of three-dimensionalcell/polymer constructs for bone tissue engineering in a spinner flask and arotating wall vessel bioreactor. J Biomed Mater Res, 62(1):136–48, 2002.

[18] E. Boyd. Mathematical modelling of the flow field and particle motion in arotating bioreactor at unit gravity and microgravity. In NASA/ASEE SummerFaculty FellowshipProgram, Johnson Space Centre, August 23 1990.

41


[19] Q. Q. Qiu, P. Ducheyne, P. S. Ayyaswamy. Fabrication, characterization andevaluation of bioceramic hollow microspheres used as microcarriers for 3-D bonetissue formation in rotating bioreactors. Biomaterials, 20(11):989–1001, 1999.

[20] S. R. Pollack, D. F. Meaney, E. M. Levine, M. Litt, E. D. Johnston. Numer-ical model and experimental validation of microcarrier motion in a rotatingbioreactor. Tissue Eng, 6(5):519–530, 2000.

[21] M. Lappa. Organic tissues in rotating bioreactors: Fluid-mechanical aspects,dynamic growth models, and morphological evolution. Biotechnol Bioeng,84(5):518–532, 2003.

[22] P. Sucosky, D. F. Osorio, J. B. Brown, G. P. Neitzel. Fluid mechanics of aspinner-flask bioreactor. Biotechnol Bioeng, 85(1):34–46, 2004.

[23] B. Bilgen, I. M. Chang-Mateu, G. A. Barabino. Characterization of mixingin a novel wavy-walled bioreactor for tissue engineering. Biotechnol Bioeng,92(7):907–19, 2005.

[24] N. Bolgen et al. Three-dimensional ingrowth of bone cells within biodegrad-able cryogel scaffolds in bioreactors at different regimes. Tissue Eng Part A,14(10):1743–50, 2008.

[25] J. L. Milan, J. A. Planell, D. Lacroix. Simulation of bone tissue formationwithin a porous scaffold under dynamic compression. Biomech Model Mechan,9(5):583–596, 2010.

[26] D. Lacroix, A. Chateau, M. P. Ginebra, J. A. Planell. Micro-finite elementmodels of bone tissue-engineering scaffolds. Biomaterials, 27(30):5326–34, 2006.

[27] A. L. Olivares, E. Marsal, J. A. Planell, D. Lacroix. Finite element studyof scaffold architecture design and culture conditions for tissue engineering.Biomaterials, 30(30):6142–9, 2009.

[28] C. Sandino and D. Lacroix. A dynamical study of the mechanical stimuli andtissue differentiation within a CaP scaffold based on micro-CT finite elementmodels. Biomech Model Mechanobiol, 10(4):565–576, 2011.

[29] C. Sandino, J. A. Planell, D. Lacroix. A finite element study of mechanicalstimuli in scaffolds for bone tissue engineering. J Biomech, 41(5):1005–14, 2008.

42


[30] J. L. Milan, J. A. Planell, D. Lacroix. Computational modelling of the mechani-cal environment of osteogenesis within a polylactic acid-calcium phosphate glassscaffold. Biomaterials, 30(25):4219–4226, 2009.

[31] E. Baas, J. H. Kuiper, Y. Yang, M. A. Wood, A. J. El Haj. In vitro bone growthresponds to local mechanical strain in three-dimensional polymer scaffolds. JBiomech, 43(4):733–9, 2010.

[32] J. Vance, S. Galley, D. F. Liu, S. W. Donahue. Mechanical stimulationof MC3T3 osteoblastic cells in a bone tissue-engineering bioreactor enhancesprostaglandin E2 release. Tissue Eng, 11(11-12):1832–9, 2005.


[34] C. Jungreuthmayer et al. A comparative study of shear stresses in collagen-glycosaminoglycan and calcium phosphate scaffolds in bone tissue-engineeringbioreactors. Tissue Eng Part A, 15(5):1141–1149, 2009.




[38] Y. Yao, Y. Yu, J. Guo, Y. Qian, J. Wang. Prediction of shear stress distributionthroughout 3D bone scaffold in perfusion environment. In Image Analysis andSignal Processing (IASP), pages 51–54, 2010.


[40] B. Porter, R. Zauel, H. Stockman, R. Guldberg, D. Fyhrie. 3-D computationalmodeling of media flow through scaffolds in a perfusion bioreactor. J Biomech,38(3):543–9, 2005.

43


[41] R. Voronov, S. VanGordon, V. I. Sikavitsas, D. V. Papavassiliou. Computa-tional modeling of flow-induced shear stresses within 3D salt-leached porousscaffolds imaged via micro-CT. J Biomech, 43(7):1279–1286, 2010.

[42] F. Maes, P. Van Ransbeeck, H. Van Oosterwyck, P. Verdonck. Modeling fluidflow through irregular scaffolds for perfusion bioreactors. Biotechnol Bioeng,103(3):621–30, 2009.

[43] M. Cioffi, F. Boschetti, M. T. Raimondi, G. Dubini. Modeling evaluation of thefluid-dynamic microenvironment in tissue-engineered constructs: A micro-CTbased model. Biotechnol Bioeng, 93(3):500–10, 2006.

[44] T. J. Lee et al. Spinner-flask culture induces redifferentiation of de-differentiatedchondrocytes. Biotechnol Lett, 33(4):829–36, 2011.

[45] E. M. Bueno, B. Bilgen, G. A. Barabino. Wavy-walled bioreactor supportsincreased cell proliferation and matrix deposition in engineered cartilage con-structs. Tissue Eng, 11(11-12):1699–709, 2005.

[46] L. Meinel et al. Engineering bone-like tissue in vitro using human bone marrowstem cells and silk scaffolds. J Biomed Mater Res A, 71(1):25–34, 2004.

[47] L. Uebersax et al. Effect of scaffold design on bone morphology in vitro. TissueEng, 12(12):3417–29, 2006.

[48] E. Baas and J. H. Kuiper. A numerical model of heterogeneous surface strainsin polymer scaffolds. J Biomech, 41(6):1374–8, 2008.

[49] H. Hagenmuller, M. Hitz, H. P. Merkle, L. Meinel, R. Muller. Design andvalidation of a novel bioreactor principle to combine online micro-computedtomography monitoring and mechanical loading in bone tissue engineering. RevSci Instrum, 81(1):014303, 2010.

[50] A. Sittichockechaiwut, A. M. Scutt, A. J. Ryan, L. F. Bonewald, G. C. Reilly.Use of rapidly mineralising osteoblasts and short periods of mechanical loadingto accelerate matrix maturation in 3D scaffolds. Bone, 44(5):822–829, 2009.

[51] D. W. Hutmacher. Scaffolds in tissue engineering bone and cartilage. Bioma-terials, 21(24):2529–43, 2000.


44


[53] G. N. Bancroft, V. I. Sikavitsas, A. G. Mikos. Design of a flow perfusion biore-actor system for bone tissue-engineering applications. Tissue Eng, 9(3):549–54,2003.

[54] M. E. Gomes, V. I. Sikavitsas, E. Behravesh, R. L. Reis, A. G. Mikos. Effectof flow perfusion on the osteogenic differentiation of bone marrow stromal cellscultured on starch-based three-dimensional scaffolds. J Biomed Mater Res A,67(1):87–95, 2003.

[55] A. M. Sailon et al. A novel flow-perfusion bioreactor supports 3D dynamic cellculture. J Biomed Biotechnol, 2009:873816, 2009.

[56] G. N. Bancroft et al. Fluid flow increases mineralized matrix deposition in 3Dperfusion culture of marrow stromal osteoblasts in a dose-dependent manner.Proc Natl Acad Sci U S A, 99(20):12600–5, 2002.

[57] D. Lacroix, J. A. Planell, P. J. Prendergast. Computer-aided design and finite-element modelling of biomaterial scaffolds for bone tissue engineering. PhilosT R Soc A, 367(1895):1993–2009, 2009.

[58] D. P. Byrne, D. Lacroix, J. A. Planell, D. J. Kelly, P. J. Prendergast. Simulationof tissue differentiation in a scaffold as a function of porosity, Young’s mod-ulus and dissolution rate: Application of mechanobiological models in tissueengineering. Biomaterials, 28(36):5544–5554, 2007.

[59] A. Boccaccio et al. Finite element method (FEM), mechanobiology andbiomimetic scaffolds in bone tissue engineering. Int J Biol Sci, 7(1):112–132,2011.

[60] J. A. Sanz-Herrera, J. M. Garcia-Aznar, M. Doblare. A mathematical modelfor bone tissue regeneration inside a specific type of scaffold. Biomech ModelMechan, 7(5):355–66, 2008.

[61] P. J. Prendergast, R. Huiskes, K. Soballe. Biophysical stimuli on cells duringtissue differentiation at implant interfaces. J Biomech, 30(6):539–548, 1997.

[62] J. Pierre and C. Oddou. Engineered bone culture in a perfusion bioreactor: a2D computational study of stationary mass and momentum transport. ComputMethods Biomech Biomed Engin, 10(6):429–38, 2007.

[63] S. L. Ishaug et al. Bone formation by three-dimensional stromal osteoblastculture in biodegradable polymer scaffolds. J Biomed Mater Res, 36(1):17–28,1997.

45


[64] B. Obradovic, J. H. Meldon, L. E. Freed, G. Vunjak-Novakovic. Glycosamino-glycan deposition in engineered cartilage: Experiments and mathematicalmodel. AIChE J, 46(9):1860–1871, 2000.

[65] J. Petrasch, P. Wyss, A. Steinfeld. Tomography-based monte carlo determina-tion of radiative properties of reticulate porous ceramics. J Quant SpectroscRadiat Transf, 105(2):180–197, 2007.

[66] E. Zermatten, S. Haussener, M. Schneebeli, A. Steinfeld. Tomography-baseddetermination of permeability and dupuit-forchheimer coefficient of character-istic snow samples. J Glaciol, 57(205):811–816, 2011.

[67] N. Datta et al. In vitro generated extracellular matrix and fluid shear stresssynergistically enhance 3D osteoblastic differentiation. P Natl Acad Sci USA,103(8):2488–93, 2006.

[68] J. Van den Dolder et al. Flow perfusion culture of marrow stromal osteoblastsin titanium fiber mesh. J Biomed Mater Res A, 64(2):235–41, 2003.

[69] V. I. Sikavitsas et al. Flow perfusion enhances the calcified matrix depositionof marrow stromal cells in biodegradable nonwoven fiber mesh scaffolds. AnnBiomed Eng, 33(1):63–70, 2005.

[70] X. Yu, E. A. Botchwey, E. M. Levine, S. R. Pollack, C. T. Laurencin. Bioreactor-based bone tissue engineering: The influence of dynamic flow on osteoblastphenotypic expression and matrix mineralization. Proc Natl Acad Sci U S A,101(31):11203–8, 2004.

[71] H. S. Tuan and D. W. Hutmacher. Application of microCT and computationmodeling in bone tissue engineering. Comput Aided Design, 37(11):1151–1161,2005.

46

Chapter 3

Development of perfusionbioreactor system

47

3.1 Perfusion bioreactor design

The following sections of this chapter describe the development of a perfusionbioreactor system for dynamic bone tissue engineering (BTE) applications with thepossibility to monitor three-dimensional (3-D) mineralized tissue formation usingmicro-computed tomography (μCT). In chapter 3.1 the design of our in-house de-signed perfusion bioreactor was adapted. This chapter is further divided into twoparts: (i) the description of the design principle of the perfusion bioreactor and (ii)the investigation of the influence of the bioreactor design on the mechanical envi-ronment inside the bioreactor and the scaffold. In chapter 3.2 the influence of thefetal bovine serum (FBS) type used on the mineralized tissue formation is described.Finally, the influcence of different flow rates applied on the mineralized tissue for-mation by human mesenchymal stem cells (hMSCs) is described in chapter 3.3.

3.1 Optimization of perfusion bioreactor design

In BTE, computational simulations have been performed over the last 20 yearsto investigate different aspects of in-vitro cultures. Simulation techniques evolvedfrom simple numerical models to complex and detailed finite element models [1].Numerous studies have been performed to investigate the effect of scaffold geometryon fluid velocity fields or shear stresses (SS) [2, 3]. The effect of the design ofthe bioreactor on the mechanical environment present in bioreactors was mostlyneglected, although it has been shown that it influences experimental outcomes [4].

In this chapter the design process of the perfusion bioreactor used for the experi-ments of this thesis is described. The design process was divided into two parts: (i)design principle of the perfusion bioreactor based on requirements for cell cultureexperiments and μCT applications and (ii) design adaptations based on computa-tional simulations performed aiming for a uniform mechanical environment in thebioreactor and the scaffold.

3.1.1 Perfusion bioreactor design principle

The perfusion bioreactor design had to fulfill various requirements. The outer biore-actor geometry had to be designed for μCT applications using a μCT 40 machine(SCANCO Medical AG, Brüttisellen, Switzerland). Conventional sample holdersfor μCT machines are provided in the form of cylindrical vials and are made fromradiolucent materials. The size of the sample holder, here the bioreactor, defines themaximal resolution that can be acquired [5]. A cylindrical outer shape of 36mm indiameter and 78mm in height was chosen for the perfusion bioreactor (Fig. 3.1.1A).These outer dimensions allowed a maximal image resolution of 18μm (high resolution

49

Chapter 3 Development of perfusion bioreactor system

mode, scan time about 1h) or lower if needed. Two openings were introduced at thetop and the bottom of the bioreactor, where the tubing was attached (Fig. 3.1.1A).So called shut-off couplings were used to connect the bioreactor to the tubing. Shut-off couplings are closed when detached and open when attached. This allows thebioreactor to be completely detached from the tubing for μCT scans. The materialused for the bioreactor had to be biocompatible and radiolucent to allow the x-raysto penetrate through the bioreactor to reach the μCT’s detector. Polysulfone meetsthese specifications and in addition it is highly resistant to acids and bases in pHranging from 2 to 13, it is resistant to oxidizing agents and exhibits a high dimen-sional stability for temperatures between -100◦C to +150◦C making the materialsuitable for sterilizing by steam autoclaving [6].

Fig. 3.1.1: (A) Bioreactor housing with the two openings for the attachment of the couplings at thetop and the bottom (*). (B) Schematic representation of the building block design of the in-housedesigned perfusion bioreactor with the bioreactor top and bottom (1, 2) and the top and bottominlay (3) fixing the scaffold (4). The fluid flow is indicated in top-bottom direction (5). (C) Top(blue) and bottom (red) inlay designed to house one compliant silk fibroin (SF) scaffold of 8mmin diameter and 3mm in height (single-scaffold design). (D) Top (blue) and bottom (red) inlaydesigned to house 8 compliant SF scaffolds of 4mm in diameter and 3mm in height (high-throughputdesign).

The bioreactor design should not only be used for perfusion cultures, but also forstatic cultures. The inner volume of the bioreactor was designed to be big enough toenclose about 6ml of culture medium, which is the same volume of medium used instatic bioreactors. The inner volume of the bioreactor was composed of the inlet andoutlet volume and two opposing cones between which the scaffold inlays were placed

50


(Fig. 3.1.1B). The perfusion bioreactor was built up from different building blocks:(1, 2) bioreactor top and bottom forming the bioreactor housing, (3) inlay top andbottom fixing the (4) scaffold (Fig. 3.1.1B). The fluid flow (5) can be applied in top-bottom direction as shown in Fig. 3.1.1B, but also in bottom-up direction accordingto the settings set at the pump. This building block principle allows the bioreactorto be adapted to various other scaffold sizes or materials. The number of scaffoldshoused in one bioreactor can also be varied to increase experimental throughput.Two different scaffold inlay types have been designed for: (i) one compliant silkfibroin (SF) scaffold of 8mm in diameter and 3mm in height (single-scaffold design;Fig. 3.1.1C) and (ii) 8 compliant SF scaffolds of 4mm in diameter and 3mm in height(high-throughput design; Fig. 3.1.1D). The advantage of an inlay housing 8 scaffoldsat once is an increase in experimental throughput. With the high-throughput designa higher number of different experimental conditions could be evaluated within thesame time compared to cultures using one scaffold in one bioreactor. The resultsof the following computational fluid dynamics (CFD) simulations were performedusing the high-throughput design. The results observed could be directly translatedto the single-scaffold design (results not shown).

3.1.2 The influence of perfusion bioreactor design on velocity

fields

The influence of the design of the perfusion bioreactor on the mechanical envi-ronment, represented as velocity fields, in the bioreactor and the scaffold was in-vestigated. Briefly, CFD were performed in COMSOL Multiphysics 4.3 (ComsolInc., Burlington, MA, United States). The three-dimensional (3-D) geometry of thebioreactor was built using computer-aided design and imported into COMSOL. Thebioreactor geometry was subsequently divided into two domains: (i) the bioreactorand (ii) the scaffold. Free (bioreactor domain) and porous (scaffold domain) mediaflow was simulated using the Free and Porous Media Flow module in COMSOL.

Simulations of the velocity fields revealed differences when different flow rateswere simulated. The velocity fields were uniform among all 8 scaffolds when a lowflow rate was simulated (corresponding to a flow rate of 0.2ml/min set at the pump).The velocity field within one scaffold was uniform and symmetric (Fig. 3.1.2A).

At a higher flow rate (12ml/min) the velocity fields among the 8 scaffolds weredifferent compared to each other and the flow field within one scaffold was inhomoge-neous and asymmetrical (Fig. 3.1.2B). It is assumed that these results are caused bythe bioreactor design. The fluid flow enters the bioreactor through the narrow inletinto the evolving cone and is then probably stopped by the scaffold inlay. This might

51


Fig. 3.1.2: Flow fields simulated in the scaffold volumes for (A) a low flow rate (correspondingto 0.2ml/min set at the pump) and (B) a high flow rate (corresponding to 12ml/min set at thepump).

cause irregularities in the velocity profile. This assumption was confirmed by lookingat the velocity field in the full bioreactor volume (Fig. 3.1.3). The original biore-actor design showed the formation of swirls above the scaffold inlay (Fig. 3.1.3A).With increasing height of the bioreactor the streamlines got more uniform, becausethere was more space for the fluid flow to settle down (Fig. 3.1.3B-D). Increasing thediameter of the inlet and outlet of the bioreactor showed a similar effect (Fig. 3.1.4).

Another option to eliminate swirls and non-uniformity in velocity fields is theapplication of flow conditioners. Flow conditioners are frequently used in the con-text of flow measurements. Inserted upstream of a flow meter, flow conditioners areable to improve the measurement accuracy of the flow rate in a pipe [7]. Differentflow conditioner geometries were designed and their effect on the velocity field inthe scaffolds was evaluated. Flow conditioners improved the flow fields in the scaf-folds vastly. Uniform and symmetrical flow fields were observed within the scaffoldvolume and among all 8 scaffolds compared to each other (Fig. 3.1.5). Flow condi-tioners produced with circular holes of 1mm diameter arranged in concentric circles(Fig. 3.1.5A) or hexagons (Fig. 3.1.5B) showed the most promising results.

3.1.3 Limitations

All three approaches (increasing bioreactor height, increasing inlet diameter, andthe insertion of flow conditioners) led to more uniform and symmetric velocity fields

52


Fig. 3.1.3: Simulations of full bioreactor volume with the velocity field indicated with streamlines(red). (A) Original bioreactor design with the height between the scaffold and the evolving cone onthe top of 5.5mm. Adapted bioreactor designs with the height increased to (B) 10mm, (C) 50mm,and (D) 100mm.

within the bioreactor and the scaffolds in the perfusion bioreactor. Nevertheless,there were some limitations. The increased bioreactor height was not feasible forμCT application due to the limitations of the outer dimensions given by the μCTmachine. Increasing the diameter of the inlet and the outlet seemed to have thelowest impact on the uniformity of the velocity fields. The original bioreactor designwas adapted for the application of flow conditioners inserted at the top and thebottom of the scaffold inlay (Fig. 3.1.6A). During the experiments performed with

53


Fig. 3.1.4: Simulations of full bioreactor volume with the velocity field indicated with streamlines(red). (A) Original bioreactor design with an inlet and outlet diameter of 2mm. Adapted bioreactordesigns with inlet and outlet diameters of (B) 5mm and (C) 7mm.

Fig. 3.1.5: (A) Simulation of velocity field in the scaffolds using a flow conditioner with circularholes of 1mm in diameter arranged in concentric circles. (B) Simulation of velocity field in thescaffolds using a flow conditioner with circular holes of 1mm in diameter arranged in concentrichexagons. Flow rate simulated: 12ml/min.

flow conditioners, major problems arose. Vast air inclusions were observed fromscout views performed right before starting the μCT scans (Fig. 3.1.6B) comparedto no air inclusions present at the start of the study (Fig. 3.1.6C). It was not possibleto get rid of the air inclusions during the study. It is assumed that air formation inthe bioreactor occurs regularly, but with the flow conditioners used the air could not

54


leave the inner volume of the bioreactor anymore. For scaffolds of 8mm in diameter itwas observed that air bubbles could penetrate through the scaffold, but not throughsmaller scaffolds of 4mm in diameter which have been used in the high-throughputperfusion bioreactor.

To avoid problems with air inclusions, the bioreactor was designed to house one SFscaffold of 8mm in diameter and 3mm in height (Fig. 3.1.1B, C). The diameter of theinlet and outlet of the bioreactor was increased to 5mm leading to a homogeneousand symmetric flow field in the bioreactor and the scaffold at higher flow rates asshown in Fig. 3.1.4B. Using this design, the throughput of the bioreactor could notbe increased, but it was more important to generate homogeneous and symmetrical

Fig. 3.1.6: (A) High-throughput perfusion bioreactor design with flow conditioners inserted aboveand below the scaffold inlays (blue). The selection corresponds to the extract shown in (B, C). (B)Scout view of high-throughput perfusion bioreactor showing vast air inclusions below and abovethe flow conditioners (indicated with *) as well as in-between the flow conditioners and the scaffoldinlays (indicated with #). (C) Scout view of high-throughput perfusion bioreactor showing no airinclusions below and above the flow conditioners (indicated with *) as well as in-between the flowconditioners and the scaffold inlays (indicated with #).

55


velocity fields in the bioreactor and the scaffolds to increase the reproducibility ofthe experiments. All experiments described in the following chapters have beenperformed with this bioreactor design.

References

[1] J. R. Vetsch, R. Müller, S. Hofmann. The evolution of simulation techniques fordynamic bone tissue engineering in bioreactors. J Tissue Eng Regen Med, 2013.



[4] E. M. Bueno, B. Bilgen, G. A. Barabino. Wavy-walled bioreactor supportsincreased cell proliferation and matrix deposition in engineered cartilage con-structs. Tissue Eng, 11(11-12):1699–709, 2005.

[5] M. Stauber and R. Muller. Micro-computed tomography: a method for the non-destructive evaluation of the three-dimensional structure of biological specimens.Methods Mol Biol, 455:273–92, 2008.

[6] D. Parker et al. Ullmann’s Encyclopedia of Industrial Chemistry, book sectionPolymers, High-Temperature. Wiley-VCH Verlag GmbH & Co. KGaA, 2000.

[7] E. M. Laws. Flow conditioning-a new development. Flow Meas Instrum,1(3):165–170, 1990.

56

3.2 Static culture conditions - fetal bovine serum

3.2 Fetal bovine serum

Effect of fetal bovine serum on mineralization in silk fibroin scaffolds

Jolanda Rita Vetsch1, Samantha Jean Paulsen1, Ralph Müller1, Sandra Hofmann1,2,3

1Institute for Biomechanics, Swiss Federal Institute of Technology Zurich (ETHZ),Vladimir-Prelog-Weg 3, Zurich 8093, Switzerland2Department of Biomedical Engineering, Eindhoven University of Technology, PO Box513, Eindhoven 5600MB, The Netherlands3Institute for Complex Molecular Systems, Eindhoven University of Technology, PO Box513, Eindhoven 5600MB, The Netherlands

published in:Acta Biomaterialia2015, 13, 277-285Postprint version according to publisher copyright policy.

Abstract:Fetal bovine serum (FBS) is a common media supplement used in tissue engineering(TE) cultures. The chemical composition of FBS is known to be highly variablebetween different brands, types or batches and can have a significant impact oncell function. This study investigated the influence of four different FBS typesin osteogenic or control medium on mineralization of acellular and cell-seededsilk fibroin (SF) scaffolds. In bone TE, mineralized tissue is considered as thefinal product of a successful cell culture. Calcium assays and micro-computedtomography scans revealed spontaneous mineralization on SF scaffolds with certainFBS types, even without cells present. In contrast, cell-mediated mineralization wasfound under osteogenic conditions only. Fourier transform infrared spectroscopyanalysis demonstrated a similar ion composition of the mineralization present inscaffolds, whether cell-mediated or spontaneous. These results were confirmed byscanning electron microscopy. This study shows clear evidence for the influenceof FBS type on mineralization on SF scaffolds. The suitability of FBS mediumsupplementation in TE studies is highly questionable with regard to reproducibilityof studies and comparability of obtained results. For future TE studies, alternativesto conventional FBS such as defined FBS or serum-free media should be considered,as suggested decades ago.

57


Keywords:Fetal bovine serum, silk fibroin, scaffolds, bone tissue engineering, spontaneous min-eralization

3.2.1 Introduction

Fetal bovine serum (FBS) is a nutritional serum supplement used for most cell cul-tures. It contains important basic proteins such as growth factors and hormones formaintaining cell survival, growth and division. The chemical composition of FBStypically varies from batch to batch, between different types and among differentbrands due to biological variance [1, 2]. FBS contains many unknown substanceswith unclear functions on cultured cells that may alter the outcome of cell experi-ments [1]. Today, various types of defined FBS are commercially available. DefinedFBS is chromatographically purified, then individual constituents are separated andrecombined into a defined composition.

Biomineralization is a process describing mechanisms of mineralized tissue for-mation by organisms in nature [2, 3]. Biominerals are hybrid structures composedof both minerals and organic components [3]. Mineral deposition starts by theformation of prenucleation clusters at a templating surface initiated by local super-saturation of ions [4, 5]. The formation of these ion clusters is highly dependent onthe fluid surrounding the underlying structure of the inorganic matrix [3, 5]. Theproteins of the inorganic matrix interact with ions of the surrounding fluid andfacilitate the formation of mineralized macromolecules [4].

The most abundant mineralized tissue in the human body is bone [4]. Bone is acomposite material consisting of hydroxyapatite (HA) and collagen type I (Col I)fibrils [6] . In-vivo bone mineralization is a cell-mediated process. Osteoblasts areresponsible for the deposition of bone matrix into the extracellular space. Col I isthe major component of bone matrix [7] and is a fibrous protein containing repetitiveamino acids (Fig. 3.2.1A). Col I builds the three-dimensional (3-D) framework ofbone on which bone mineral is deposited [3]. The mineralization process in boneoccurs by nucleation of HA out of calcium and phosphate ions in solution. Theorganic Col I acts as a substrate for the mineralization of HA crystals that mineralizein thin layers between sequences of Col I molecules (Fig. 3.2.1A) [8].

Like Col I, silk fibroin (SF) is a fibrous protein. It is synthesized by Lepidopteralarvae [9]. SF protein side chains interdigitate and form antiparallel plated β-sheets(Fig. 3.2.1B). Highly and less ordered β-sheets are connected by amorphous net-

58


Fig. 3.2.1: Comparison of collagen and silk fibroin structures. Both fibrous proteins show a hierar-chical structure with repeating amino acids and have been reported to take part in mineralizationprocesses. (A) Structure of collagen with repeating amino acids of glycine (Gly), proline (Pro) andhydroxyproline (Hyp). (B) Structure of silk fibroin with repeating amino acids of glycine (Gly),serine (Ser) and alanine (Ala).

work chains [10]. SF is a widely used scaffold material for bone tissue engineering(TE) applications [11–15], due to its excellent biocompatibility [16], controllabledegradation [17] and favorable mechanical properties [18]. SF is an interesting scaf-fold material for bone TE considering its ability to regulate the formation of HAnanocrystals when exposed to simulated body fluid (SBF). SBF mimics ion concen-trations of human blood serum [19, 20]. It was shown that SF has the potentialto mineralize spontaneously in SBF by inducing apatite deposition on its surface

59


and provoking continuous growth and enrichment of HA crystals [20,21]. Marelli etal. [10] have shown that the amorphous connections between the β-sheets of SF actas nucleation sites for HA crystals similar to Col I in bone. Spontaneous depositionof mineralization on SF was also shown in calcium chloride (CaCl2) solution. Choiet al. [23] managed to promote calcium deficient HA formation on SF particles inCaCl2 solution, due to electrostatic interactions between the calcium ions and thefunctional groups of SF [22]. Ion compositions in FBS and SBF are very similar andthe pH, an important environmental factor influencing spontaneous mineralization,of both solutions is buffered to 7.4 [19,23].

In bone TE applications, the formation of mineralized extracellular matrix bycells seeded on a 3-D scaffold is considered as the final product of a successfulculture. Still, obtained results vary highly within and between research groupsand reproducibility of studies is often not given. If FBS type (or even batch) hasan influence on mineralization due to variations in chemical composition betweendifferent suppliers, this effect needs to be quantified or preferably avoided in orderto draw conclusions on how the cellular environment influences TE outcomes. Thereare some studies which investigated the effect of FBS on mineralization focusing onthe cellular response, but without having a closer look at the role of the materialsused [24, 25]. The objective of this paper was to investigate the influence of fourdifferent FBS types on mineralized tissue formation in cell-seeded and acellular SFscaffolds. Two conventional and two defined FBS types were compared againsteach other. Additionally, the influence of osteogenic factor supplementation onmineralized tissue formation was evaluated.

3.2.2 Materials and methods

Details of all FBS types used (full label, order number, batch number and detailedsupplier information), all experimental conditions (FBS type, medium type andcell use) and all assays performed (including number of samples evaluated for eachexperimental condition) can be found in Table 3.2.1.

60


Table 3.2.1: Supplier information of each FBS type, experimental conditions with group numbersindicated and assays performed including numbers of samples evaluated for each experimentalcondition per time point.

PAA Gold PAA Standard BiochromSuperior

Gibco Standard

Full label FBS Gold, EUapproved

PAA FBS standardquality, EUapproved

Biochrom FBSSuperior

Gibco R©, FBS,Qualified, EUapproved, SouthAmerica Origin

Abbreviation PAA Gold PAA Standard Biochrom Superior Gibco StandardOrdernumber

A15–151 A15–101 S 0615 10270–106

Batchnumber

A15110–2039 A10110–2069 0306 A 41Q6208 K

Supplier Chemie Brunschwig,Basel, Switzerland

Chemie Brunschwig,Basel, Switzerland

Biochrom AG,Berlin, Germany

Gibco, Zug,Switzerland

ExperimentalconditionsN=44 pergroup

1control acellular2osteo acellular3control cell-seeded4osteo cell-seeded




Assaysperformed

week 3:Calcium assay(N=10)

week 5:Calcium assay(N=10)

week 7:Calcium assay(N=10)μCT(N=10)FTIR (N=10)SEM (N=10)

Materials

The four different FBS types used were chosen according to the following criteria.Gibco Standard is the standard FBS currently used in our lab. Like Gibco Standard,PAA Standard is a conventional FBS and was chosen as a direct control for GibcoStandard. PAA Gold and Biochrom Superior are both defined FBS types and werechosen because the suppliers assert no need for batch-to-batch testing. Dulbecco’smodified Eagle medium (DMEM), antibiotic/antimycotic (Anti-Anti) and trypsinwere from Gibco (Zug, Switzerland). 1,1,1,3,3,3-hexafluoroisopropanol (HFIP) wasobtained from abcr chemicals (Karlsruhe, Germany). Methanol (MeOH) was fromMerck (Zug, Switzerland) and lithium bromide (LiBr) from Thermo Fisher Scien-tific (Reinach, Switzerland). Phosphate buffered saline (PBS) was supplied fromMedicago (Uppsala, Sweden). All other substances were of analytical or pharma-ceutical grade and obtained from Sigma (Buchs, Switzerland). Silkworm cocoonswere kindly provided by Trudel Inc. (Zurich, Switzerland).

61


Scaffolds

SF scaffolds were produced as previously described [13,26]. Briefly, silk from Bombyxmori L. silkworm cocoons was degummed by boiling in 0.2M Na2CO3 twice for 1h.Dried silk was dissolved in 9M LiBr and dialyzed against ultra pure water (UPW)for 36h using Slide-A-Lyzer cassettes (molecular weight cutoff: 3.5K; Thermo FisherScientific, Reinach, Switzerland). Dialyzed silk solution was lyophilized (Alpha 1–2,Martin Christ GmbH, Osterode am Harz, Germany) for 4 days and dissolved inHFIP, resulting in a 17% (w/v) solution. 1ml of dissolved silk was added to 2.5gNaCl with a granule size of 300–400μm and was allowed to air dry for 3 days. Silk-salt blocks were immersed in 90% MeOH for 30min to induce β-sheet formation [27].NaCl was extracted from dried blocks in deionized water for 2 days. Scaffolds werecut into disks of 5mm in diameter and 3mm in height and autoclaved in PBS at121◦C for 20min.

Cell Culture

Human mesenchymal stem cell (hMSC) isolation from human bone marrow (Lonza,Walkersville, MD, USA) was performed as previously described [28]. Passage 3 hM-SCs were expanded for 7 days. At day 7 hMSCs were trypsinized and half of allscaffolds were seeded with 1 million cells per scaffold, while the remaining scaffoldswere left acellular. All scaffolds were incubated in wells of a 24-well plate at 37◦C for90min to allow cell attachment. Subsequently, half of all cell-seeded scaffolds andhalf of all acellular scaffolds were provided with 1ml control medium (DMEM supple-mented with 10% FBS and 1% Anti-Anti) and were incubated at 37◦C and 5% CO2.The remaining scaffolds were incubated in 1ml osteogenic medium (control mediumsupplemented with 50μgml−1 L-ascorbic acid 2-phosphate, 100nM dexamethasoneand 10 mM β-glycerophosphate (βGP)). Every culture medium was prepared withevery type of FBS, resulting in 16 experimental conditions (Table 3.2.1).

Calcium Assay

A spectrophotometric calcium assay was performed according to the manufacturer’sinstruction (Calcium CPC FS, Rolf Greiner BioChemica, Flacht, Germany). Af-ter 3, 5 and 7 weeks of culture, 10 scaffolds per group were washed with PBSand disintegrated in 5% aqueous trichloroacetic acid (TCA) using steel balls anda MinibeadBeaterTM(Biospec, Bartlesville, OK, USA). Scaffolds were further incu-bated in 5% TCA at room temperature for 48h and frozen until evaluation.

For the correlation of calcium assay and micro-computed tomography (μCT) data,

62


20 scaffolds were analyzed after μCT scanning for calcium amount. After the μCTscans, scaffolds were disintegrated and incubated as described before. For analysis,all scaffolds were centrifuged at 3000g for 10min. Supernatant was further diluted in5% TCA at a ratio of 1:20 (v/v) and the color intensity of the resulting solution wasmeasured with a microplate reader (Infinite 200 PRO, Tecan Group Ltd, Männedorf,Switzerland).

μCT

μCT scans were performed on scaffolds after 7 weeks of culture (N=10 per group).Scaffolds were washed with PBS, lyophilized overnight (Alpha 1–2, Martin ChristGmbH, Osterode am Harz, Germany) and scanned in air on a μCT 40 (ScancoMedical, Brütisellen, Switzerland). The energy level was set to 45kVp, intensity to177μA, integration time to 200ms and a frame averaging of 4 was applied. The scanswere executed at medium resolution mode with a nominal resolution of 12μm. Afterreconstruction, scans were filtered applying a 3-D constrained Gaussian filter withfinite filter support (1 voxel) and filter width (sigma=1.2). Filtered grayscale imageswere segmented at a global threshold of 17% of the maximal grayscale value (corre-sponding to a density value of 125.17mgHAcm-3). Unconnected objects smaller than50 voxels were removed and neglected for further analysis. For the final evaluationan overall mask (OM) with radius=Ro and height=H was generated (Fig. 3.2.2A).Ro and H were chosen to fit to the outer boundaries of each scaffold individuallydue to small differences in scaffold dimensions especially present between mineral-ized and non-mineralized scaffolds. The resulting 3-D volume was evaluated mor-phometrically for mineralized tissue volume (BV) and mineralized tissue volumefraction (BV/TV=mineralized tissue volume/total volume), as described previouslyfor bone [29, 30]. All scaffolds were grouped into either an “acellular” or a “cell-seeded” group to assess differences in spatial distribution of spontaneously formedmineralization on acellular scaffolds and of cell-mediated mineralization observed oncell-seeded scaffolds. An inner mask (IM) was generated based on the geometry ofthe OM (Fig. 3.2.2B). The percentage of total BV within the IM was calculatedbased on the BV of the OM. The radius (Ri) of the IM equaled 50%*Ro. H was keptconstant (Fig. 3.2.2B). Bone mineral content (BMC) per scaffold as determined byμCT was correlated with calcium amount per scaffold as measured biochemically.

Fourier transform infrared spectroscopy (FTIR)

After 7 weeks of culture, scaffolds (N=2 per group) were washed with PBS, frozenat -80◦C overnight and lyophilized for 1 day. For FTIR analysis dried scaffolds were

63


Fig. 3.2.2: Masks applied for the evaluation of μCT data. (A) OM. The OM was fitted to thewhole volume of each scaffold individually. An outer scaffold radius (Ro) and the scaffold height(H) were chosen to fit to the dimensions of each scaffold. (B) IM. The IM mask was created basedon the OM with a radius of Ri=50%*Ro. The height (H) of both masks was kept constant.

ground with KBr at 4% (w/w). 100% KBr was used as background medium. Theground samples were analyzed using a Bruker Tensor 27 FTIR spectroscope (BrukerOptics GmbH, Fällanden, Switzerland) at a resolution of 4cm−1. Scans were per-formed in full range from 4000cm−1 to 400cm−1 at a scan time of 100ms. Measure-ments were converted to absorbance spectra in OPUS Application Executable 5.0.5(Bruker Optics GmbH, Fällanden, Switzerland). SigmaPlot 12.2 (Systat Software,Inc., San Jose, CA, USA) was used to determine exact wavenumbers for peaks andAdobe Illustrator CS5.1 (Adobe Systems, Inc., San Jose, CA, USA) was used tomark peaks.

Scanning electron microscopy (SEM)

After 7 weeks of culture, scaffolds (N=2 per group) were washed in PBS and fixed in2.5% glutaraldehyde solution in 0.1M cacodylate buffer (pH 7.4) at 4◦C for 4h. Afterrinsing in 0.1M cacodylate buffer a second fixation step was performed using 0.04%aqueous osmium tetroxide in 0.1M cacodylate buffer at room temperature in thedark. Scaffolds were again rinsed in 0.1M cacodylate buffer, dehydrated for 10minin serial EtOH baths of 37%, 67% and three times 96%, frozen overnight in 100%EtOH at -80◦C and lyophilized. Scaffolds were coated with gold using a sputtercoater (Balzers SCD 030, former Balzers Union Ltd, Liechtenstein) for 90–120s at acurrent of 40mA and imaged at 850x and 5000x magnification in an SEM (Leo 1530

64


Gemini, Zeiss, Oberkochen, Germany) operated at a voltage of 5kV to investigatescaffold surface morphology.

Statistical analysis

Quantitative data are represented as mean±standard deviation. IBM SPSS Statis-tics 20 (SPSS Inc., Chicago, IL, USA) was used for the evaluation of statisticallysignificant differences for calcium assay and quantitative μCT data. Under the as-sumption of normally distributed data, analysis of variance was performed followedby post hoc assessment using the Bonferroni method. Differences between groupswere considered statistically significant at a level of P�0.05. Images in Fig. 3.2.4represent the upper median BV/TV sample of each group.

3.2.3 Results

Calcium Assay

Calcium assays were performed to assess total calcium content of the scaffolds, asa method to represent the total mineralized tissue volume per scaffold (Fig. 3.2.3A-C). As expected, cells responded to the different culture media used. Cells seededon scaffolds cultured in control medium remained undifferentiated, whereas cellsseeded on scaffolds in osteogenic medium underwent osteogenic differentiation andstarted to form mineralized tissue. Scaffolds cultured in osteogenic medium exhib-ited the highest calcium contents after 7 weeks (Fig. 3.2.3C). Interestingly, consider-able amounts of calcium were detected for acellular scaffolds with certain FBS typesstarting as early as week 3 up to week 7. Elevated calcium levels were observed par-ticularly for acellular scaffolds cultured with Biochrom Superior and Gibco Standard(Fig. 3.2.3A-C).

Cell-seeded scaffolds cultured in osteogenic medium showed significantly increasedcalcium contents compared to cell-seeded scaffolds cultured in control medium andall acellular scaffolds at each time point (P�0.05). No differences were observedbetween cell-seeded scaffolds cultured in control medium at any time point.

Acellular scaffolds cultured in osteogenic medium with Gibco Standard showedsignificantly increased calcium deposition compared to acellular and cell-seeded scaf-folds cultured in control medium with the same FBS type at week 3 (P�0.05;Fig. 3.2.3A). In week 5 a similar result was observed for acellular scaffolds cul-tured in control and osteogenic medium with Biochrom Superior with significantlyhigher calcium levels compared to their corresponding cell-seeded scaffolds culturedin control medium (P�0.05; Fig. 3.2.3B). The same result was observed for acellular

65


scaffolds cultured in control and osteogenic medium with Gibco Standard at week7 (P�0.05; Fig. 3.2.3C).

Fig. 3.2.3: Calcium assay and μCT data. Calcium amount per scaffold in μg after: (A) 3 weeks,(B) 5 weeks and (C) 7 weeks of culture. (D) Mineralized tissue volume fraction after 7 weeks of cul-ture. Lines represent significant differences between FBS types under the same culture conditions.abcrepresent significant differences from control acellular, osteogenic acellular and control cell-seeded conditions, respectively, within the same type of FBS. Error bars represent mean±standarddeviation. Results are considered statistically significant at P�0.05. N=10 per group for calciumassay and μCT data.

μCT

μCT measurements of BV/TV (Fig. 3.2.3D) confirmed calcium assay data of week 7(Fig. 3.2.3C). Cell-seeded scaffolds cultured in osteogenic medium exhibited thehighest BV/TV values compared to cell-seeded scaffolds cultured in control mediumand all acellular scaffolds (P�0.001). Acellular scaffolds cultured with BiochromSuperior and Gibco Standard showed significantly increased calcium levels comparedto their corresponding cell-seeded scaffolds cultured in control medium (P�0.05).

Mineralized tissue was located primarily on scaffold edges and the top of the scaf-folds in cell-seeded scaffolds. Acellular scaffolds cultured with Biochrom Superiorand Gibco Standard showed substantial mineralization (Fig. 3.2.4A and B). Very

66


Fig. 3.2.4: 3-D μCT images of scaffold mineralization for each FBS type after 7 weeks of culture.The upper median sample of each group is shown in top view (upper row) and side view (bottomrow): (A) control acellular, (B) osteogenic acellular, (C) control cell-seeded and (D) osteogeniccell-seeded conditions. Scale bar=1cm.

little mineralization was observed in acellular scaffolds cultured with PAA Gold orPAA Standard or in cell-seeded scaffolds cultured in control medium (Fig. 3.2.4A-C). Cell-seeded scaffolds cultured in osteogenic medium displayed highly mineralizedstructures with every type of FBS (Fig. 3.2.4D). For the evaluation of differencesin spatial distribution of acellular scaffolds and cell-seeded scaffolds only mineral-ized scaffolds (BV of OM�0.001mm3) were considered. Acellular scaffolds (N=77)

67


showed significantly increased BV percentage in the IM (18.87±18.32%) comparedto cell-seeded scaffolds (N=52) that displayed 10.66±11.16% of the total BV in theIM (P�0.05; Fig. 3.2.5). BMC per scaffold correlated highly with calcium amountper scaffold (R2=0.96; Supplementary Fig. S3.2.1A). Both measurement methodswere in agreement and no systematic differences could be observed between calciumassay and μCT data (Bland Altman Plot; Supplementary Fig. S3.2.1B).

Fig. 3.2.5: BV in percentage of total BV per scaffold for the IM volume after 7 weeks of culture.Samples with a BV<0.001mm3 were excluded from evaluation leading to N=77 for the acellularand N=52 for the cell-seeded condition. Higher BV was observed in the IM for acellular scaffolds.Error bars represent mean±standard deviation. Results are considered statistically significant atP�0.05.

FTIR

FTIR was performed to analyze differences in phosphate and carbonate precipitateson the scaffolds (Fig. 3.2.6; Supplementary Table S3.2.1). Typical absorption spec-tra of the underlying SF scaffold were observed between 1700 and 1000cm−1 for everyculture condition [31]. All spectra showed features of different apatite precipitates.Carbonate was present in all samples [10, 32] and phosphate bands were visiblebetween 1200 and 900cm−1 [33–36]. Cell-seeded scaffolds cultured in osteogenicmedium displayed a characteristic peak for phosphate ions (PO3−

4 ) at 1038cm−1,indicating a higher amount of mineralized tissue formation (Fig. 3.2.6D) [10, 33].Overlapping phosphate bands between 1200 and 900cm−1 present in spectra of acel-lular scaffolds might have masked less prominent phosphate peaks.

68


Fig. 3.2.6: FTIR data: (A) control acellular, (B) control cell-seeded, (C) osteogenic acellular and(D) osteogenic cell-seeded conditions. Spectra are shown between wavenumbers of 1700–800cm−1.

SEM

SEM micrographs confirmed the results of the calcium assay and μCT data(Fig. 3.2.7; Supplementary Fig. S3.2.2). Cell-seeded scaffolds showed extracellularmatrix covering the surface of the scaffolds (Supplementary Fig. S3.2.2I-P). No min-eralization clusters were visible on acellular scaffolds cultured in osteogenic mediumscaffolds with PAA Gold (Fig. 3.2.7A and C). In contrast, clearly visible mineraliza-tion clusters were observed with Gibco Standard under the same culture conditions(Fig. 3.2.7B and D).

69


Fig. 3.2.7: SEM images of acellular scaffolds cultured in osteogenic medium. (A, B) Overviewpictures of one characteristic scaffold per group at a magnification of 850x and (C, D) close-uppictures of the area marked in the respective pictures in (A) or (B) at a magnification of 5000x.Mineralized clusters are present on the surface of scaffolds cultured with Gibco Standard. Nomineralized clusters and a smooth scaffold surface can be observed for scaffolds cultured with PAAGold. Scale bar: 60μm (A, B) and 10μm (C, D).

3.2.4 Discussion

FBS has been used for more than a century as a cell culture additive for most celltypes cultured in-vitro [37, 38]. It has been previously shown that differences inserum composition lead to statistical differences in experimental outcomes [39]. Thedisadvantages of FBS, such as undefined composition or batch-to-batch variations,have been known for decades. Alternatives to FBS have been proposed and includeoptions such as sera from other species, reduced serum media, chemically definedFBS or chemically defined serum-free media [37,38]. Using defined FBS would elim-inate the necessity to test every new batch of FBS and would minimize experimentaldifferences due to different FBS batches or types. Nevertheless, if correct batch test-ing is performed, experimental differences could already be reduced. It is importantthat every FBS batch is tested, preferably under real culture conditions.

70


Very few publications can be found discussing differences in experimental out-comes provoked by FBS. The exact ordering and batch number of the FBS usedin experiments is not required for publications and therefore a direct comparisonof experiments from different laboratories seems impossible. It is very difficult toassess whether experimental outcomes occur due to the experimental conditions orif they are attributed to the FBS type or batch. Differences in FBS compositionmay be a major reason why different labs are not able to reproduce data publishedin the literature.

In this study, the effect of four different FBS types on mineralized tissue formationon SF scaffolds was investigated. The influence of FBS on mineralization is not newand has been addressed earlier. For example, mineral precipitation on decalcifiednewborn rat tibia in the absence of cells has been shown to occur when culture mediawere supplemented with FBS and osteogenic factors (βGP) [25], but not with FBS-containing control media only as shown with this study. Evident effects of FBS onmineralization, alkaline phosphatase activity and osteogenic marker expression basedon serum conditions have also been shown for human adipose stem cells, but theresults were not compared to acellular culture conditions and therefore the influenceof the material could not be determined [24]. The results observed in the presentstudy revealed that the FBS type influences mineralization formation on acellularand cell-seeded SF scaffolds significantly. The addition of osteogenic factors did nothave an influence on mineralized tissue formation in acellular scaffolds, but led toosteogenic differentiation with cell-seeded scaffolds.

Acellular scaffolds showed an unexpected and completely different mineraliza-tion pattern compared to cell-seeded scaffolds. Scaffolds cultured with BiochromSuperior and Gibco Standard show strong evidence of spontaneous mineralizationon acellular SF scaffolds presenting calcium values up to one third of the highestcalcium level measured overall. The amount of spontaneous mineralization of acel-lular scaffolds was therefore clearly dependent on the FBS type used. Interestingly,spontaneous mineralization occurred even under control conditions. This result iscontradictory to previous results where mineral precipitation in cell-free conditionsonly occurred with FBS containing media supplemented with βGP [25].

The amount of spontaneous mineralization is dependent on the ion concentrationof the solution surrounding the substrate [20]. Differences in spontaneous mineral-ization between the four FBS types used can be explained by differences in elementalcomposition. It is known that sera exhibit significant differences in composition be-tween different types and batches [39, 40]. The elemental composition of FBS, orsera in general, is not defined. Only some common components are defined for all

71


FBS types (Supplementary Table S3.2.2). Basic properties like pH or osmolality,primary components like Hemoglobin or proteins and the absence of bacteria, fungiand mycoplasma were tested for all the four FBS types used. The ion concentrationwas stated for Biochrom Superior only (Supplementary Table S3.2.2). Hamlin andPrice [25] hypothesized that there exist one or more nucleators of bone mineraliza-tion in serum that escaped from the bone matrix to the blood. For future studies itwould be helpful to identify key components that contribute to mineralization.

Cell-seeded scaffolds showed mineralization when cultured with osteogenicmedium only (Fig. 3.2.3 and Fig. 3.2.4). It is known that hMSCs differenti-ate towards the osteogenic lineage with the addition of osteogenic factors to themedium [41]. The presence of mineralization in these scaffolds shows a successfuldifferentiation of the hMSCs towards osteoblasts. The amount of mineralizationformed was dependent on the FBS type, especially at early time points (week 3 and5; Fig. 3.2.3A and B), which is in agreement with previous results [24].

In cell-seeded scaffolds, we assume that cells prevented spontaneous mineralizationby either changing the ion concentration of the surrounding culture medium or byinhibiting the interaction between ions in the culture medium with either the SFsurface itself or the proteins adsorbed to the surface of the scaffold. Under osteogenicconditions cells seem to metabolize the ions in the culture medium to build upmineralized tissue.

The formation of spontaneous mineralization under control conditions on acel-lular scaffolds compared to cell-seeded scaffolds, however, is not fully understood.One possibility is that cells cultured with control medium take up ions from thesurrounding medium like cells cultured under osteogenic conditions. In contrast,cells subsequently are not able to form mineralized tissue, because they are not dif-ferentiated towards the osteogenic lineage. Another possible mechanism inhibitingspontaneous mineralization in cell-seeded scaffolds under control conditions couldbe an active inhibition of mineralization by the hMSCs itself. It is known that somecell types are able to inhibit mineralization [42]. So far, active inhibition of mineral-ization by hMSCs could not be shown and further investigation would be needed totest this hypothesis. Spontaneous mineralization of acellular SF has been describedbefore in SBF or CaCl2 [20–22] but, to our knowledge, it was never shown in controlor osteogenic medium or on other scaffold materials.

The underlying silk structure is thought to provide an ideal environment forspontaneous mineralization, because its chemical structure is very similar to Col I(Fig. 3.2.1). Comparisons of mineralization on different scaffold materials are there-fore critical, because some materials could induce spontaneous mineralization more

72


effectively than others because of their different chemical structure. Therefore, itis necessary to include acellular control groups to test spontaneous mineralizationpotential of each scaffold material with the respective culture media used.

It is known that in static cell cultures seeded cells tend to concentrate on theouter scaffold surface, leading to a poor nutrient and waste exchange to the centerof the scaffold [43]. Because of this effect the cell-mediated mineralization is ex-pected to be located predominantly at the outer scaffold surface in the presentedstudy setup. This effect was confirmed by μCT scans showing mineralized tissueformation located mostly at the edges and the top of the scaffolds (Fig. 3.2.4D). Ionsare small enough to diffuse throughout the whole scaffold volume easily. Becausespontaneous mineralization is dependent on local ion concentration [20], the spatialdistribution of spontaneous mineralization on acellular scaffolds was expected to bemore homogeneous compared to cell-seeded scaffolds. Quantitative evaluation ofspatial distribution of mineralization volume confirmed this hypothesis and showeda significantly increased percentage of total BV in the center of acellular scaffolds(Fig. 3.2.5).

SEM provided additional visual confirmation of the results detected with thecalcium assay and μCT (Supplementary Fig. S3.2.2). The absence of cells in acellularscaffolds was confirmed, and indicated that the mineralization in acellular scaffoldsis purely attributable to the spontaneous mineralization phenomena.

FTIR revealed that the mineralization formed in all of the 16 experimental groups,whether cell-mediated or not, consisted of similar phosphate and carbonate precip-itates. Unfortunately, FTIR lacks the possibility to clearly distinguish between alldifferent precipitates present, especially when they are available in small amounts.FTIR gives information about the chemical composition of the samples only. In ad-dition to FTIR, X-ray diffraction could be used to determine crystallinity (size andorientation) to compare the mineralization formed to in-vivo bone mineralization.In addition to that, transmission electron microscopy can be performed to investi-gate the orientation of mineral crystals relative to the orientation of collagen fibrils.To characterize mineral mechanical properties of the formed mineralization atomicforce microscopy can be used [44]. The solubility behavior of a mineral phase couldreveal further information about the composition of the solid phase [45].

With respect to clinical applications it is of great importance that the effectsof FBS on cell cultures are known and animal components are avoided. Humanserum is an interesting alternative to animal-derived serum. It seems attractivefor the culture of tissue-engineered grafts for future implantation and clinical celltherapy applications. Because of the clinical relevance of human serum, it would be

73


interesting to investigate its effect on the mineralization on acellular and cell-seededSF scaffolds. Nevertheless, results that have been observed with human serum arestill conflicting. The use of human serum, however, makes sense only if enoughautologous serum is available. Due to the limited availability and high variabilitywithin different donors, human serum is not a reasonable option for large-scale cellcultures in clinical applications [46] and therefore it is highly recommended to lookfor serum-free alternatives for in-vitro cultures.

The use of conventional FBS for cell culture is questionable and should be consid-ered with reservation. Not only cell growth, but also cell viability, cell attachment,cell density, geometrical environment (2-D/3-D) and scaffold material should beconsidered for every FBS type used [37]. Chemically defined FBS, like PAA Gold,could prevent excessive and time-consuming batch-to-batch testing. Ideally, the useof defined FBS is to be preferred over conventional FBS.

It should be noted that the results of this paper are valid only for the describedexample of bone TE with hMSCs and 3-D SF scaffolds prepared by the methodsdescribed herein. However, it can be assumed that similar results could be expectedfor any other output parameter in any other type of static cell culture. The ef-fect of different FBS types in dynamic cell cultures needs to be investigated moreprecisely. Results are expected to be different for dynamic cultures because thescaffold-surrounding medium is moving, which could highly influence total ion con-centration in the scaffold and therefore the process of spontaneous mineralization.This study is also limited to the four different FBS types used and batch variationsfor each individual FBS type were not investigated. It has to be expected that otherFBS types or batches show different mineralization amounts and patterns on SFscaffolds.

It is important that these differences will be revealed in the near future. Hopefully,cell culture will soon be performed using serum-free medium if possible. If, for anyreason, serum-free medium cannot be used, suitable FBS pre-testing needs to beperformed and it should be a matter to state at least FBS type and batch usedin publications. Alternatively, standards for batch testing procedures for specificapplications should be determined, similar to cell characterization protocols. Ifthese recommendations are followed in the future, it is possible that the complexityof TE cultures can be reduced, leading to more robust and more comparable results.

3.2.5 Conclusions

The objective of this study was to investigate the effect of different FBS types onTE cultures; in this example, on bone TE cultures of hMSCs on 3-D SF scaffolds.

74


The results provide clear evidence of the strong influence of FBS type on mineraliza-tion in SF scaffolds using different evaluation methods. Unexpectedly, spontaneousmineralization was observed in acellular scaffolds dependent on FBS type. Only twoof the four FBS types used induced spontaneous mineralization under acellular con-ditions. The addition of osteogenic factors to the medium did not influence sponta-neous mineralization. No mineralization was present in cell-seeded scaffolds culturedwith control medium, which indicates that cells are involved in the prevention ofspontaneous mineralization. The observed results emphasize a clear dependence ofexperimental outcomes on FBS type. The use of FBS in TE cultures is thereforehighly questionable and alternatives to conventional FBS should be considered, asalready suggested decades ago. It is important to solve these issues now in orderto reduce variability and improve experimental comparability to promote the entirefield of TE.

Conflict of Interest

The authors declare that there is no conflict of interest.

Acknowledgements

The authors would like to acknowledge Trudel Inc. (Zurich, Switzerland) for thekind supply of silkworm cocoons. We thank Dr. Dirk Mohn for his help in FTIR dataacquisition and evaluation. This project was supported by the European Union (EUProject No. FP7-NMP-2010-LARGE-4: BIODESIGN—Rational Bioactive Materi-als Design for Tissue Regeneration).

References

[1] C. E. A. Jochems, J. B. F. van der Valk, F. R. Stafleu, V. Baumans. The useof fetal bovine serum: Ethical or scientific problem? Atla-Altern Lab Anim,30(2):219–227, 2002.

[2] M. Sarikaya. Biomimetics: materials fabrication through biology. Proc NatlAcad Sci U S A, 96(25):14183–5, 1999.

[3] S. Weiner and P. M. Dove. An overview of biomineralization processes and theproblem of the vital effect. Biomineralization, 54:1–29, 2003.

[4] L. Addadi and S. Weiner. Control and design principles in biological mineral-ization. Angew Chem Int Ed Engl, 31(2):153–169, 1992.

75


[5] A. Dey et al. The role of prenucleation clusters in surface-induced calciumphosphate crystallization. Nat Mater, 9(12):1010–4, 2010.

[6] J. Y. Rho, L. Kuhn-Spearing, P. Zioupos. Mechanical properties and the hier-archical structure of bone. Med Eng Phys, 20(2):92–102, 1998.

[7] E. J. Mackie. Osteoblasts: novel roles in orchestration of skeletal architecture.Int J Biochem Cell Biol, 35(9):1301–5, 2003.

[8] P. Fratzl, N. Fratzl-Zelman, K. Klaushofer, G. Vogl, K. Koller. Nucleationand growth of mineral crystals in bone studied by small-angle x-ray scattering.Calcif Tissue Int, 48(6):407–413, 1991.

[9] D. Kaplan, S. Mello, S. Arcidiacono, S. Fossey, K. Senecal. Protein basedmaterials, book Silk. Birkhäuser, Boston, 1998.

[10] B. Marelli et al. Silk fibroin derived polypeptide-induced biomineralization ofcollagen. Biomaterials, 33(1):102–8, 2012.

[11] H. Hagenmuller, M. Hitz, H. P. Merkle, L. Meinel, R. Muller. Design andvalidation of a novel bioreactor principle to combine online micro-computedtomography monitoring and mechanical loading in bone tissue engineering. RevSci Instrum, 81(1):014303, 2010.

[12] L. Meinel et al. Silk implants for the healing of critical size bone defects. Bone,37(5):688–98, 2005.

[13] L. Meinel et al. Bone tissue engineering using human mesenchymal stem cells:Effects of scaffold material and medium flow. Ann Biomed Eng, 32(1):112–122,2004.

[14] S. Sofia, M. B. McCarthy, G. Gronowicz, D. L. Kaplan. Functionalized silk-based biomaterials for bone formation. J Biomed Mater Res, 54(1):139–148,2001.

[15] L. Uebersax et al. Effect of scaffold design on bone morphology in vitro. TissueEng, 12(12):3417–29, 2006.

[16] G. H. Altman et al. Silk matrix for tissue engineered anterior cruciate ligaments.Biomaterials, 23(20):4131–41, 2002.

[17] T. Arai, G. Freddi, R. Innocenti, M. Tsukada. Biodegradation of bombyx morisilk fibroin fibers and films. J Appl Polym Sci, 91(4):2383–2390, 2004.

76


[18] J. Perez-Rigueiro, C. Viney, J. Llorca, M. Elices. Mechanical properties ofsingle-brin silkworm silk. J Appl Polym Sci, 75(10):1270–1277, 2000.

[19] T. Kokubo and H. Takadama. How useful is SBF in predicting in vivo bonebioactivity? Biomaterials, 27(15):2907–2915, 2006.

[20] A. Takeuchi et al. Deposition of bone-like apatite on silk fiber in a solutionthat mimics extracellular fluid. J Biomed Mater Res A, 65(2):283–9, 2003.

[21] A. S. Lin, T. H. Barrows, S. H. Cartmell, R. E. Guldberg. Microarchitecturaland mechanical characterization of oriented porous polymer scaffolds. Bioma-terials, 24(3):481–9, 2003.

[22] Y. Choi et al. Silk fibroin particles as templates for mineralization of calcium-deficient hydroxyapatite. J Biomed Mater Res B, 100(8):2029–2034, 2012.

[23] T. Lindl. Zell- und Gewebekultur. Spektrum Akademischer Verlag, Heidelberg,5th edition, 2002.

[24] L. Kyllonen et al. Effects of different serum conditions on osteogenic differen-tiation of human adipose stem cells in vitro. Stem Cell Res Ther, 4:17, 2013.

[25] N. J. Hamlin and P. A. Price. Mineralization of decalcified bone occurs undercell culture conditions and requires bovine serum but not cells. Calcified TissueInt, 75(3):231–242, 2004.

[26] R. Nazarov, H. J. Jin, D. L. Kaplan. Porous 3-D scaffolds from regenerated silkfibroin. Biomacromolecules, 5(3):718–26, 2004.

[27] M. Tsukada et al. Structural-changes of silk fibroin membranes induced byimmersion in methanol aqueous-solutions. J Polym Sci Pol Phys, 32(5):961–968, 1994.

[28] S. Hofmann et al. Control of in vitro tissue-engineered bone-like structuresusing human mesenchymal stem cells and porous silk scaffolds. Biomaterials,28(6):1152–62, 2007.


77


[30] G. H. van Lenthe et al. Nondestructive micro-computed tomography for biologi-cal imaging and quantification of scaffold-bone interaction in vivo. Biomaterials,28(15):2479–90, 2007.

[31] D. Wilson, R. Valluzzi, D. Kaplan. Conformational transitions in model silkpeptides. Biophys J, 78(5):2690–701, 2000.

[32] S. Hossain et al. Fabrication and intracellular delivery of doxorubicin/carbonateapatite nanocomposites: effect on growth retardation of established colon tu-mor. PLoS One, 8(4):e60428, 2013.

[33] M. C. Chang and J. Tanaka. FT-IR study for hydroxyapatite/collagennanocomposite cross-linked by glutaraldehyde. Biomaterials, 23(24):4811–8,2002.

[34] A. Stoch et al. FTIR absorption-reflection study of biomimetic growth of phos-phates on titanium implants. J Mol Struct, 555:375–382, 2000.

[35] C. Drouet. Apatite formation: why it may not work as planned, and how toconclusively identify apatite compounds. Biomed Res Int, 2013:490946, 2013.

[36] C. Li, C. Vepari, H. J. Jin, H. J. Kim, D. L. Kaplan. Electrospun silk-BMP-2scaffolds for bone tissue engineering. Biomaterials, 27(16):3115–24, 2006.

[37] D. W. Jayme, D. A. Epstein, D. R. Conrad. Fetal bovine serum alternatives.Nature, 334(6182):547–8, 1988.

[38] D. Brunner et al. Serum-free cell culture: the serum-free media interactiveonline database. ALTEX, 27(1):53–62, 2010.

[39] N. Bryan, K. D. Andrews, M. J. Loughran, N. P. Rhodes, J. A. Hunt. Elu-cidating the contribution of the elemental composition of fetal calf serum toantigenic expression of primary human umbilical-vein endothelial cells in vitro.Biosci Rep, 31(3):199–210, 2011.

[40] P. J. Price and E. A. Gregory. Relationship between invitro growth promotionand biophysical and biochemical-properties of the serum supplement. In VitroCell Dev B, 18(6):576–584, 1982.

[41] M. F. Pittenger et al. Multilineage potential of adult human mesenchymal stemcells. Science, 284(5411):143–7, 1999.

78


[42] T. Oshima et al. Myeloma cells suppress bone formation by secreting a solubleWnt inhibitor, sFRP-2. Blood, 106(9):3160–5, 2005.


[44] A. L. Boskey and R. Roy. Cell culture systems for studies of bone and toothmineralization. Chem Rev, 108(11):4716–33, 2008.

[45] F. C. M. Driessens, J. W. E. Vandijk, J. M. P. M. Borggreven. Biologicalcalcium phosphates and their role in the physiology of bone and dental tissuesI. composition and solubility of calcium phosphates. Calc Tiss Res, 26(2):127–137, 1978.

[46] K. Bieback et al. Human alternatives to fetal bovine serum for the expansionof mesenchymal stromal cells from bone marrow. Stem Cells, 27(9):2331–41,2009.

79


Supplementary Data

Supplementary Figures

Fig. S3.2.1: Correlation of calcium assay data (calcium amount=CA) per scaffold with micro-computed tomography (μCT) data (BMC=bone mineral content) per scaffold: (A) Linear regres-sion showing the correlation between calcium assay and μCT data (R2=0.96, y=0.14x). (B) BlandAltman plot comparing calcium assay and μCT data. The plot shows that both evaluation methodsare in agreement and no systematic differences are present.

80


Fig. S3.2.2: Scanning electron microscopy (SEM) pictures: (A-D) Control acellular, (E-H) os-teogenic acellular, (I-L) control cell-seeded, and (M-P) osteogenic cell-seeded conditions. Smallimages: Overview pictures of one characteristic scaffold per group at a magnification of 850x;scalebar=120μm. Big images: Close-up pictures of the respective small image at a magnificationof 5000x; scalebar=10μm.

81


Supplementary Tables

Table S3.2.1: Fourier transformed infrared spectroscopy (FTIR) data. Peak assignments for allpeaks observed in FTIR spectra (Supplementary Fig. S3.2.2).

Wavenumber Peak Assignements Comments Litearture1669–1655a

1636–1626b1◦ amide, C=O stretching,silk I peaks

Weaker peaks in acellular scaffolds, verystrong peaks in cell-seeded scaffolds

[33]

1541–1526a 2◦ amide, N-H deformation,amide II band

All peaks very strong, cell-seeded scaffoldsshow slightly shifted peaks (+10cm−1)

[33]

1450–1447 N-CH3, amides Strong peaks in all groups, carbonate [10] [32]

1408–1406 N-CH3, amides Strong peaks in all groups, carbonate [10]

1265–1258b

1238–1233a3◦ amide, N-H deformation Strong peaks in all groups [33]

1166–1152 HPO4 Strong peaks in all groups [33] [36]

1076–1072 P=O stretching, HPO4 Strong peaks in all groups [33] [35]

1038 O-H stretching, PO3−4 Very strong peak only in cell-seeded

scaffolds with osteogenic medium[10] [33]

aattributed to random coil structure of silk fibroin; battributed to β-sheet conformation of silk fibroin

Table S3.2.2: Complete list of all components and properties listed in the certificates of analysis(CoA) of the four fetal bovine serum (FBS) types used. Components and properties tested for allFBS are highlighted in grey. Values in brackets display the specification ranges for each componentor property for each FBS type.

PAA Gold PAAStandard

BiochromSuperior

GibcoStandard

Appearance, color typical typical clear, amber satisfactorySterility ok ok no microbial

growthnegative

pH [-] 7.1 (6.8–8.2) 7.2 (6.8–8.2) 7.26 (6.40–8.20) 7.00 (6.90–7.80)Osmolality [mOsmol kg/H2O] 306 (280–340) 314 (280–340) 325 (280–340) 312 (280–340)Hemoglobin [mg/dl] 14.4 (�20) 16.5 (�20) 26.05 (<50) 15.20

(1.00–30.00)Endotoxin [EU/ml] <0.5 (�30) 2.62 4.64 n/aGrowth promotion 75% (�75%) 87% (�75%) complies with

controln/a

Cloning efficiency n/a n/a complies withcontrol

1.00 (0.70–1.40)

Identification DRID n/a n/a exclusivelybovine origin

n/a

BVDC/MVC not detected not detected not detected testedIBRV/BHV-1 not detected not detected not detected negativePIV 3 not detected not detected not detected negativeBVDV/MVD (virus against) n/a n/a 2 n/aBHV-1 antibody n/a n/a <2 n/aBVD-1 antibody not detected not detected n/a n/aBVD-2 antibody not detected not detected n/a n/aPIV 3 antibody n/a n/a <2 n/a

82


Table S3.2.2 continuedPAA Gold PAA

StandardBiochromSuperior

GibcoStandard

Mycoplasma not detected not detected not detected not detectedALP [U/l] n/a n/a 367 n/aGOT (AST) [U/l] n/a n/a 52 n/aGPT (ALT) [U/l] n/a n/a 11 n/agamma-GT [U/l] n/a n/a 5 n/aBilirubin [μmol/l] n/a n/a 1.71 n/aLDH [U/l] n/a n/a 569 n/aCK (NAC) [U/l] n/a n/a 138 n/aCholesterol [mmol/l] n/a n/a 0.75 n/aTriglyceride [mmol/l] n/a n/a 0.59 n/aCreatinine [μmol/l] n/a n/a 271.4 n/aUrea [mmol/l] n/a n/a 5.16 n/aSodium [mmol/l] n/a n/a 139 n/aPotassium [mmol/l] n/a n/a 11.3 n/aCalcium [mmol/l] n/a n/a 3.75 n/aMagnesium [mmol/l] n/a n/a 1.26 n/aPhosphate [mmol/l] n/a n/a 2.61 n/aIron [μmol/l] n/a n/a 30.8 n/aGlucose [mmol/l] n/a n/a 7.55 n/aProtein total [g/100 ml] 4.2 (3.0–4.5) 3.7 (3.0–4.5) 3.3 3.95 (3.50–5.50)IgG [μg/ml] 40 (�100) 94 (�250) 20 n/aAlbumine [g/100 ml] 2.5 2.0 1.52 2.22 (1.70–3.50)alpha-Globulin [g/100 ml] 0.9 0.9 1.59 1.4 (0.70–2.00)beta-Gloulin [g/100 ml] 0.7 0.7 0.15 0.32 (0.30–0.90)gamma-Globulin [g/100 ml] 0.1 0.1 40 0.14247

(0.01000–0.20000)Estradiol [pg/ml] n/a n/a 16.3 n/aProgesterone [ng/ml] n/a n/a <0.20 n/aTestosterone [ng/ml] n/a n/a 0.02 n/aElectro Profile n/a n/a n/a typicalDiploid Growth n/a n/a n/a 0.90 (0.70–1.40)Plating Efficiency n/a n/a n/a 0.98 (0.70–1.40)Toxicity n/a n/a n/a satisfactoryRel Myel/hybrid n/a n/a n/a 0.93 (0.70–1.40)Abbreviations: EU=endotoxin units; DRID=drug-related infectious disease; BVDV=bovine virusdiarrhoe-virus; MVC=canine minute virus; IBRV=bovine rhinotracheitis virus; BHV-1/2=bovine herpes virus type1/type 2; PIV 3=parainfluenza virus type 3; ALP=alkaline phosphatase; GOT (AST)=glutamic-oxaloacetictransaminase (aspartat-aminotransferase); GPT=glutamic-pyruvic transaminase; gamma-GT=gamma-glutamyl transpeptidase; LDH=actic acid dehydrogenase; CK (NAC)=creatine kinase (N-acetylcysteine);IgG=immunoglobulin G; PE=plating efficiency.

83


3.3 Flow velocity

Flow velocity-driven differentiation of human mesenchymal stem cells in silk fibroinscaffolds: A combined experimental and computational approach

Jolanda Rita Vetsch1, Duncan Betts1, Ralph Müller1, Sandra Hofmann1,2,3


in preparation

Abstract:

Mechanical loading plays a major role in bone remodeling and fracture healing.Mimicking the concept of mechanical loading of bone has been widely studiedin bone tissue engineering (BTE) by perfusion cultures. Nevertheless, there isstill debate regarding the in-vitro mechanical stimulation regime. This studyaims at investigating the effect of two different flow rates (vlow=0.001m/s andvhigh=0.061m/s) on the mineralized tissue growth of human mesenchymal stem cells(hMSCs) cultured on 3-D silk fibroin scaffolds. The two flow rates applied werechosen to mimic mechanical loading during early fracture healing or during boneremodeling, respectively. Computational fluid dynamics simulations were performedto assess SS acting on the cultured hMSCs subjected to fluid flow. Time-lapsedmicro-computed tomography (μCT) demonstrated mineralized extracellular matrixformation at vhigh only. Biochemical assays and histology confirmed these resultsand revealed increased cell proliferation at vlow. Visual mapping of simulated SS to3-D μCT data revealed SS between 0.06mPa to 0.39mPa to induce cell proliferationand SS between 0.55mPa and 24mPa to induce osteogenic differentiation of hMSCs.This study showed the feasibility to drive cell behavior of hMSCs by the flowvelocity applied in agreement with mechanical loading during fracture healing(vlow) or during bone remodeling (vhigh). These results can be used in the fu-ture to tightly control cell behavior of hMSCs towards proliferation or differentiation.

84

3.3 Dynamic culture conditions - flow velocity

Keywords:Bone tissue engineering, computational fluid dynamics, shear stress, mesenchymalstem cell, ECM (extracellular matrix)

3.3.1 Introduction

Mechanical loading plays an important role in the remodeling of mineralized bonematrix and fracture healing. In healthy bone, the mineralized bone matrix is con-tinuously remodeled, as a function of local mechanical stimuli [1]. In adult bone,osteocytes are considered to be the mechanosensitive cell. They are thought to senseshear stresses (SS) caused by load-induced movement of the interstitial fluid withinthe lacuno-canalicular system [2]. Overstimulation of osteocytes by high SS leads tothe recruitment of osteoblasts to the bone surface and subsequent mineralized bonematrix formation [3]. During fracture healing on the other hand, bone is formed by acascade of events like gradual stiffening of the forming tissue and tissue deformation.Cells within the repair tissue experience fluid flow as a consequence of loading butthe physiological effect of fluid flow is different compared to the effect of fluid flowon osteocytes. Still, the precise role of mechanical stimulation on cells in fracturehealing is not clearly understood [4].

In bone tissue engineering (BTE), researchers have been trying to mimic theconcept of mechanical loading of bone cells by perfusion flow cultures that are con-sidered to represent the loading concept the closest [5]. Perfusion cultures have beenshown to enhance osteogenic differentiation in mouse osteoblast precursor cells andrat bone marrow stromal cells [6–11]. Human derived stem cells cultured underperfusion showed increased levels of osteogenic markers and mineralized matrix de-position [12–16]. Despite these positive effects, it has also been shown that someperfusion culture settings did not support osteogenic differentiation. Very high per-fusion velocities led to apoptosis [7, 17], whereas low flow velocities increased cellproliferation in preference to osteogenic differentiation [7,13], and some studies evenshowed chondrogenic differentiation [18].

At present, a debate exists regarding the role of mechanical stimulation in in-vitro BTE cultures. In contrast to mimicking the mechanical loading present inhealthy bone, it may be more reasonable to apply a mechanical loading regime thatmimics new bone growth and repair. Osteocytes were observed starting from day 5of fracture healing, but with very short and irregularly distributed canaliculi [19].Based on these observations, it is assumed that precursor cells play a major role

85


in early fracture healing. Preferably, cultures mimicking fracture healing should beperformed with precursor cells [4]. Precursor cells, like human mesenchymal stemcells (hMSCs), have been widely used in the field of BTE [20]. The big advantage ofhMSCs is their large proliferation potential, which makes them more suitable for ex-pansion in-vitro, compared to differentiated cells [4]. Their multipotency and abilityto differentiate towards the osteogenic lineage makes hMSCs an attractive choice forclinical applications for bone regeneration [21]. Perfusion studies performed withhMSCs showed osteogenic differentiation of hMSCs [13,16], but very little is knownabout the loading regime leading to this differentiation.

The mechanical stimulation of cells in-vitro depends not only on the mechanicalloading regime applied, but also on the scaffold material used. Especially in per-fusion cultures, structural parameters of the scaffold like porosity or permeabilitycan have a significant influence on the experimental outcome [22]. Silk fibroin (SF)scaffolds have been widely used for BTE applications [23, 24] due to their excellentbiocompatibility [25] and favorable mechanical properties [26]. SF scaffolds havebeen applied successfully for bone regeneration in-vivo [27, 28] and they have beencultured with hMSCs to engineer bone-like tissue in-vitro [24].

The aim of this study was twofold: First, the effect of two different flow velocities(vlow=0.001m/s and vhigh=0.061m/s) on the behavior of hMSCs cultured on SF scaf-folds in a perfusion bioreactor was investigated. The two flow velocities have beenchosen to mimic mechanical loading during early fracture healing and mechanicalloading during bone remodeling in healthy bone tissue, respectively. We hypoth-esized that vlow leads to cell proliferation representing the early stage of fracturehealing [4], whereas vhigh leads to osteogenic differentiation of the hMSCs. Sec-ondly, SS in SF scaffolds were modeled to estimate the mechanical forces acting onthe hMSCs in various locations within the scaffold volume. Both porosity and per-meability of the SF scaffolds have been considered in the simulations. The resultingSS were subsequently mapped to the sites of mineralized extracellular matrix for-mation (ECM) assessed by micro-computed tomography (μCT) scans of the in-vitrocell cultures. We hypothesized that different SS can be defined to lead to differentcell behavior for hMSCs. In the future, SS levels found in this study are intendedto be used for BTE application to drive hMSC behavior towards the specific aim ofthe experiment.

86


3.3.2 Methods

Materials

Dulbecco’s Modified Eagle Medium (DMEM), fetal bovine serum (FBS; ordernumber 10270–06), penicillin-streptomycin-fungizone (P/S/F), nonessential aminoacids (NEAA), basic fibroblast growth factor (bFGF), β-glycerolphosphate (βGP),ascorbic acid (AA), dexamethasone (Dex), alamarBlue R© solution and Quant-iTTM

PicoGreen R© double stranded DNA (dsDNA) reagent kit were from Gibco (Zug,Switzerland). 1,1,1,3,3,3-hexafluoroisopropanol (HFIP) was from abcr GmbH&Co.(Karlsruhe, Germany). Methanol (MeOH) was from Merck (Zug, Switzerland) andLithium Bromide (LiBr) from Thermo Fisher Scientific (Reinach, Switzerland). Allother substances were of analytical grade and were purchased from Sigma (Buchs,Switzerland). Silkworm cocoons were kindly supplied by Trudel Silk Inc (Zurich,Switzerland).

Scaffolds

SF scaffolds were prepared as described earlier [23,29]. Briefly, silk cocoons from B.mori silkworm were boiled twice for 1h in 0.02M Na2CO3 and rinsed with ultra purewater (UPW). The silk was dissolved in 9M LiBr and dialyzed against UPW (Slide-A-Lyzer 3.5K MWCO, Thermo Fisher Scientific, Waltham, MA, United States) for36h, lyophilized for 4 days and dissolved subsequently in HFIP. The resulting 17%(w/v) silk solution was added to 2.5g NaCl of 315–400μm granule size and the HFIPwas allowed to evaporate for 3 days. Silk-salt blocks were immersed into 90% MeOHfor 30min to induce β-sheet formation [30]. Blocks were air dried over night andNaCl was extracted by immersing in UPW for 2 days. Disc-shaped scaffolds, 8mmin diameter and 3mm in height, were prepared using a razor blade and a biopsypunch. Scaffolds were autoclaved submerged in phosphate buffered saline (PBS) at121◦C for 20min.

Cell culture and scaffold seeding

hMSCs (Lonza, Walkersville, MD, United States) from human bone marrow aspiratewere isolated and characterized as described before [31]. P3 hMSCs were expanded inexpansion medium (DMEM, 10% FBS, 1% P/S/F, 1% NEAA and 1ng/ml bFGF)under standard cell culturing conditions (37◦C, 5% CO2) for 7 days until about80% confluence. Cells were resuspended in control medium (DMEM, 10% FBS, 1%P/S/F) at a concentration of 100 million cells per 1ml. 5 million cells were seeded oneach scaffold by adding 50μl cell suspension by pipetting. Scaffolds were incubated

87


in a 12-well plate for 90min in an incubator. Then, 1ml control medium was addedto each well and cells were allowed to attach to the scaffold for 24h. The followingday the scaffolds were transferred into the bioreactors.

Bioreactor culture

Cell seeded scaffolds were cultured in in-house designed perfusion bioreactors(Fig. 3.3.1A). Bioreactors were provided with 12ml of osteogenic medium (controlmedium, 10mM βGP, 50μg/ml AA, 100nM Dex). The medium was replaced 3 timesa week by removing 6ml medium and replacing it by double concentrated osteogenicmedium (control medium, 20mM βGP, 100μg/ml AA, 200nM Dex) to keep the ini-tial concentration of osteogenic factors constant. The bioreactors were divided intotwo different groups: (1) low perfusion and (2) high perfusion (N=5 per group). Theflow rates were set at the pump: Qlow=0.2ml/min for the low perfusion group andQhigh=12ml/min for the high perfusion group, respectively. The bioreactor culturewas maintained for 40 days.

μCT monitoring

Time-lapsed μCT images of all samples (N=5 per group) were taken once a weekover the last 5 weeks of the bioreactor culture (weeks 2 to 6) to monitor 3-D min-eralized ECM formation as described before [32]. Samples were scanned in a μCT80 (SCANCO Medical AG, Brüttisellen, Switzerland) at a voxel resolution of 36μm.Energy level was set to 45kVp and an intensity of 177μA. An integration time of200ms and 2-fold frame averaging were chosen. Gaussian filtration using a filterwidth of 1.2 and support of 1 was performed to reduce noise. Mineralized ECM wassegmented by thresholding at 97.5mg/cm3 hydroxyapatite (corresponding to a grey-scale value of 12.7%). Unconnected particles smaller than 50 voxels were removedfrom further evaluation using component labeling. The resulting 3-D volume wasevaluated morphometrically for mineralized ECM volume (BV) and volume fraction(BV/TV=mineralized ECM volume/total volume), as described previously [33,34].

Cell metabolic activity

Cell metabolic activity was tested using the alamarBlue R© assay at culture week 6.Every scaffold (N=3 per group) was incubated in 1ml of a 10% (v/v) alamarBlue R©

solution in control medium for 150min. Fluorescence of supernatant was read intriplicates at an excitation wavelength of 560nm and emission wavelength of 590nmusing a plate reader (Tecan Group Ltd., Männedorf, Switzerland). The fluorescence

88


value of each sample was normalized to the sample’s corresponding cell number.

Osteogenic differentiation

Osteogenic differentiation of the hMSCs was assessed using a colorimetric ALP assay.Scaffolds (N=3 per group) were washed with PBS, incubated in 1ml 0.2% (v/v)Triton-X-100 solution in an aqueous 5mM MgCl2 solution and disintegrated witha Mini-BeadbeaterTM (BioSpec Products Inc., Bartlesville, OK, United States) onice. After centrifugation at 3000g for 10min the ALP assay was performed withthe supernatant. The absorbance was read at 405nm using a plate reader (TecanGroup Ltd., Männedorf, Switzerland) and the amount of p-nitrophenol per samplewas calculated according to the values of a p-nitrophenol standard curve. The ALPvalue of each sample was normalized to the sample’s corresponding DNA value.Scaffolds were then incubated at room temperature for 48h, before they were usedfor DNA quantification.

DNA quantification

The Quant-iTTM PicoGreen R© dsDNA reagent kit was used for DNA quantification.After centrifugation, the DNA assay was performed according to the manufacturer’sinstructions using the sample supernatant. Fluorescence was read at an excitationwavelength of 480nm and an emission wavelength of 520nm with a plate reader(Tecan Group Ltd., Männedorf, Switzerland). The amount of DNA per sample wascalculated according to the values of a DNA standard curve.

Histology

Scaffolds (N=2 per group) were fixed in 10% (v/v) normal buffered formalin overnight at 4◦C and embedded in paraffin. Vertical cross-sections through the middleof the scaffolds were cut to a thickness of 5μm. Hematoxylin&Eosin (H&E) stainingwas performed to visualize cell nuclei and ECM. Von Kossa (VK) staining wasperformed to visualize mineralized ECM. Briefly, sections were incubated in a 1%silver nitrate solution (w/w in UPW) and photochemically degraded to silver byexposing to UV light for 45min. The sections were fixed in a 5% silver thiosulfatesolution (w/w in UPW) for 2min, dried over night and mounted the following day.

Computational modeling

Computational fluid dynamics (CFD) analyses were performed using the FPMFmodule (Free and Porous Media Flow) in COMSOL Multiphysics 4.3b (Comsol

89


Inc., Burlington, MA, United States). The geometry was built from computer-aided drawings (CAD) of the perfusion bioreactor used for the cell experiments(Fig. 3.3.1A; SolidWorks 2013, Waltham, MA, United States). The 3-D geometry ofthe perfusion bioreactor was inverted (Fig. 3.3.1B) and subsequently imported intoCOMSOL. The geometry was divided into two different domains: (1) the bioreactor(free media flow) and the (2) scaffold (porous media flow; Fig. 3.3.1C). The bioreac-tor was meshed using the built-in mesher optimized for fluid dynamics simulation.A mesh sensitivity study was performed. The convergence criterion applied was de-fined such that the relative difference of the mean velocity between two consecutiverefinement steps was below 2% in the two domains. The scaffold domain was mod-eled according to Zermatten et al. [35] with a porosity of 55%. The permeability ofthe scaffold was determined according to the method of Ochoa et al. [36] and was setto 1.76*10−11m2. No-slip boundary conditions were applied and the fluid was mod-eled as water at 37◦C. Two different inlet velocities were modeled: vlow=0.001m/sand vhigh=0.061m/s corresponding to the low and high flow rates set respectivelyat the pump during the bioreactor culture. Simulation results were correlated with3-D images from μCT scans of week 6 of the cell culture by performing ReceiverOperating Characteristic (ROC). The μCT images were superimposed on the CFDresults; SS were identified for every voxel in each scaffold and separated for min-eralized and non-mineralized ECM. All scaffolds were processed in a single ROCanalysis. By incrementally thresholding SS it was possible to determine the numberof correctly classified voxels as mineralized (true positive rate) and the number offalsely classified voxels as mineralized (false positive rate). The ROC curve wascalculated for vhigh only, because no mineralization was formed at vlow.

Fig. 3.3.1: (A) 3-D computer-aided design model of in-house designed perfusion bioreactor. (B) In-verted volume of perfusion bioreactor. (C) Material definitions for computational fluid dynamicsmodel. (1) Bioreactor: free media flow; (2) Scaffold: porous media flow. Scale bar=1 cm.

90


Statistics

Data was statistically evaluated using PASW Statistics 20.0 (SPSS Inc., ChicagoIL, United States). All quantitative data is presented as means±standard deviation(SD). Student’s t-test was performed for unpaired and paired data. Comparisonsof more than two means were done by an analysis of variance followed by Bonfer-roni post-hoc corrections. Data was considered statistically significant at p<0.05.ROC curves were calculated using an in-house developed script in Matlab R2014a(MathWorks, Natick, MA, United States).

3.3.3 Results

μCT monitoring

None of the samples of the vlow group showed any BV formation after 6 weeks of cul-ture (Fig. 3.3.2A). BV formation was observed for the vhigh group only (Fig. 3.3.2B).BV growth was observed starting from week 3 of culture, growing from the edgesof the scaffold towards the middle of the scaffold (Fig. 3.3.3B). BV/TV in the vhigh

group increased over time from 0.05%±0.03% at week 2 to 0.64%±0.20% at week 6of culture (p<0.001; Fig. 3.3.3A). After 6 weeks of culture BV/TV of the vhigh groupwas significantly higher than the BV/TV of the vlow group (p<0.001; Fig. 3.3.3A).

Cell metabolic activity

Cell metabolic activity was equal for both perfusion groups with a trend towardslower cell metabolic activity in the vlow group (p=0.19; Fig. 3.3.4A).

Osteogenic differentiation

ALP activity was higher in the high perfusion group compared to ALP activity inthe vlow group (p<0.05; Fig. 3.3.4B).

DNA quantification

The amount of DNA per sample was equal for both perfusion groups with a trendtowards higher cell number in the vlow group (p=0.10; Fig. 3.3.4C).

Histology

H&E staining showed a uniform distribution of cells and ECM throughout the wholescaffold thickness for the vlow group (Fig. 3.3.5A), whereas cells and ECM in the vhigh

91


Fig. 3.3.2: Three-dimensional reconstructed images from time-lapsed micro-computed tomography(μCT) scans. (A) Images of all samples of low perfusion group at week 6 of the culture. (B) Imagesof all samples of high perfusion group at week 6 of the culture. Scale bar: 1cm.

Fig. 3.3.3: (A) Bone-like tissue volume fraction (BV/TV) of week 2 to 6 of the cell culture. Thegrowth of bone-like tissue was initiated at week 3 of the cell culture in the high perfusion group.BV/TV of the high perfusion group increased over time and was higher at week 6 compared withthe low perfusion group (p<0.001). (B) Images of one representative sample (sample 01) of highperfusion group shown at each scan time-point. BV grew from the edge towards the middle of thescaffold. Scale bar: 1cm.

92


Fig. 3.3.4: Biochemical assay data. (A) Cell viability normalized to DNA amount. (B) Osteogenicdifferentiation (represented in alkaline phosphatase (ALP) activity) normalized to DNA amount.(C) DNA amount. *p<0.05.

93


group were more located towards the bottom of the scaffold volume (Fig. 3.3.5B).VK staining revealed no mineralization on scaffolds of the vlow group (Fig. 3.3.5C),but mineralized ECM was observed on scaffolds of the vhigh group (Fig. 3.3.5D).

Fig. 3.3.5: Histology pictures of vertical cross-sections through the middle of the scaffold. (A,B) Hematoxylin&Eosin (H&E) staining. (C, D) Von Kossa (VK) staining. Cells and extracellularmatrix (ECM) were uniformly distributed throughout the whole scaffold volume in the low perfu-sion group (A) compared to the high perfusion group (B) where cells and ECM were more locatedtowards the bottom of the scaffold. No VK staining was observed in the low perfusion group (C)whereas mineralized nodules were observed in the high perfusion group (D). Scale bar: 500μm.

Computational modeling

The velocity field through the whole bioreactor volume is visualized color-codedfor one vertical cross-section through the middle of the bioreactor in Fig. 3.3.6A.The highest velocities were observed at the inlet and outlet of the bioreactor. Themaximal velocity in the bioreactor at vlow was 1.97*10−3m/s and 89.23*10−3m/s atvhigh. The maximal flow velocity in the scaffold was 0.07*10−3m/s and 4.47*10−3m/sat vlow and vhigh, respectively. Maximal SS in the scaffold volume were 0.56mPaat vlow and 34.20mPa at vhigh. Horizontal and vertical (Fig. 3.3.6B) cross-sectionsthrough the middle of the scaffold revealed that the highest SS occurred close to the

94


bioreactor wall. SS in these cross-sections ranged from 0.06mPa to 0.39mPa at vlow

and from 0.40mPa to 24mPa at vhigh and did not overlap.

Fig. 3.3.6: Velocity and shear stress (SS) simulation data. (A) Velocity fields of vertical cross-sections at the middle of the bioreactor at low flow rate (left) and high flow rate (right). (B) SSfields of horizontal (blue) and vertical (red) cross-sections through the middle of the scaffold atlow flow rate (left) and high flow rate (right). Highest velocities were observed at the inlet andthe outlet of the bioreactor and highest SS values were observed close to the bioreactor wall in thehigh perfusion group.

Visual mapping of SS to 3-D reconstructed images from μCT scans of week 6showed that mineralized ECM volume was only formed when cells had been sub-jected to SS in a range of 0.55mPa to 24mPa (Fig. 3.3.7A).

The ROC analysis revealed that the prediction of mineralized ECM is distinctfrom random with an area under the curve of 0.69. The least random SS (furthestpoint from the 45◦ line) for predicting mineralization was 1.47mPa, with a truepositive rate of 0.81 and a false positive rate of 0.50 (Fig. 3.3.7B).

3.3.4 Discussion

It has been shown that perfusion-induced SS increased osteogenic marker expressionand mineralized matrix deposition in human derived stem cells [12–16]. A manifold-ness of SS can be found in the literature to induce osteogenic differentiation ofstem cells, but to date it is still not completely understood how SS affect stem cellbehavior exactly. The two different flow velocities applied in the study presentedhave been chosen to mimic SS occurring during early fracture healing and bone re-modeling in healthy bone tissue, respectively. A dose-dependent differentiation of

95


Fig. 3.3.7: (A) Visual mapping of 3-D micro-computed tomography (μCT) data to simulated shearstresses (SS). Mineralized extracellular matrix volume was only observed when cells have beensubjected to a SS of at least 0.55mPa. (B) Receiver Operating Characteristic (ROC) curve forsuperimposition of 3-D μCT images at week 6 of culture with simulated SS. The curve shows thatthe SS simulation is distinct from random (random corresponds to 45◦ line) and the least randomSS was observed at 1.47mPa with a true positive rate of 0.81 and a false positive rate of 0.50.

hMSCs on 3-D matrices in a perfusion setup at flow velocities between 0.1ml/minand 1.5ml/min has been shown before [13]. Nevertheless, the observed results in thecurrent study were surprisingly distinct. μCT monitoring showed no formation ofmineralized ECM at vlow at all, whereas mineralized ECM formation was observedat vhigh (Fig. 3.3.2). The results observed by μCT monitoring were confirmed bybiochemical assays and histology (Fig. 3.3.4; Fig. 3.3.5). These findings reveal aclear dependence of hMSC behavior on the perfusion velocity applied. hMSCs cul-tured at vlow did not differentiate towards the osteogenic lineage although they havebeen subjected to osteogenic medium. During fracture healing progenitor cells in-vading the repair tissue are thought to be exposed to mechanical loads stimulatingcell proliferation and matrix production [4]. The low flow velocity applied mightbe able to mimic these mechanical loads of early fracture healing leading to theincreased proliferation and matrix production observed. From the results displayed,it is not possible to confirm this theory, but it could be tested in future experiments

96


by looking for specific markers of fracture healing.

Cartmell et al. [7] observed similar results. Low flow velocity increased cellularproliferation rate whereas high flow velocity upregulated the osteogenic differentia-tion potential after 20 days of culture. The comparison to results of other perfusionstudies is nevertheless very difficult. Due to the different bioreactor and scaffoldgeometries used it is difficult to directly compare perfusion velocities and corre-sponding SS. The CFD method presented here can serve as a platform to reducethese limitations. The geometry of the bioreactor, porosity and permeability of thescaffold can be included in CFD simulations. SS can then be directly comparedbetween single studies leading to a reduced variation between the different perfusionstudies.

Mineralized ECM growth started from the edges and evolved towards the middle ofthe scaffolds with time. The application of μCT monitoring reveals distinct changesin the very same sample enabling to study the influence of mechanical loading moreclosely. Based on these results, it can be assumed that the mechanical environmentwithin the scaffold volume changed over time due to ECM deposition and led to thisgrowth pattern. At the onset of the culture, high SS occurred close to the bioreactorwall (Fig. 3.3.6B). Mineralized ECM started to grow close to the bioreactor wallfilling up the pores of the scaffold. Due to the closing of these pores, high SS tend tomove towards the middle of the scaffold. Cells cultured closer to the scaffold middlewill then be subjected to higher SS leading to subsequent mineralized ECM growthcloser to the scaffold middle.

Histology revealed that, in scaffolds cultured at vhigh, cells and ECM were prefer-ably located towards the bottom of the scaffold. It is hypothesized that the higherfluid velocity together with gravity forced cells to move towards the bottom of thescaffold. This effect could probably be prevented in future studies by applying oscil-latory flow. Cell metabolic activity and cell number was not significantly differentfor vlow and vhigh showing that the high flow velocity did not have a detrimentaleffect on the hMSCs.

CFD is a useful tool to compute SS in perfused scaffold structures. A varietyof simulation techniques have been used from simple analytical to very complexcomputational models including for example μCT based scaffold structures [37–39]. Several perfusion studies estimated the mechanical loading regime using thecylindrical pore model [7, 8, 37]. However, Jungreuthmayer et al. [38] showed thatthe model overestimates SS, especially at higher flow velocities. The advantage ofthe applied simulation model is that it includes the porosity and the permeability ofthe scaffold used and is still very low in computational costs (less than 1h to solve the

97


model) compared to simulations of real scaffold geometries. Nevertheless, to improvethe accuracy of our simulation model the real geometry of the SF scaffolds usedshould be included. μCT based simulation of SS in scaffolds for tissue engineeringhas been performed before [38–41]. These studies show very exact calculations of SSbut are mostly limited to a scaffold sub-volume due to limitations in computationalcosts. Zermatten et al. [41] have been able to simulate SS in a SF scaffold within thesame bioreactor geometry as used for this study, but the major drawback of theirstudy was the very high computational cost (several days). Another limitation ofμCT-based simulations is that most scaffolds (polymers, gels etc.) are not visible inμCT scans when immersed in culture medium, because they take up the liquid andthen their density is not distinguishable from the culture medium. Therefore, thereal geometry of the scaffold in culture medium cannot be assessed for simulationsand has to be approximated [41,42].

CFD simulations showed no overlap of SS present at vlow and vhigh. Given ourdata, maximal SS at vlow can be considered too small for the induction of osteogenicdifferentiation of hMCSs. Visual mapping of SS to 3-D μCT images from week 6of the cell culture points towards optimal SS ranging from 0.55mPa to 24mPa forosteogenic differentiation of hMSCs (Fig. 3.3.7A). The ROC analysis showed thatthere is a quantifiable link between areas of high SS and mineralized ECM. Thevolume of the SF scaffold was not taken into account in the ROC analysis becauseit could not be distinguished from the culture medium in the μCT scans due to itslow density. This then led to an artificially high number of non-mineralized voxels.This effect is small as the scaffolds are highly porous (about 90%), but could haveincreased the number of false positives in the ROC analysis by the volume fractionof the scaffold phase.

SS that induced mineral deposition in this study are similar to values observedin other in-vitro cultures [9, 13, 42]. Increased mineral deposition of bone marrowstromal cells was observed for a SS ranging from 10mPa to 30mPa compared tosamples cultured at SS smaller than 10mPa or static samples [9]. Similarly, prolifer-ation and cell metabolic activity of osteoblasts were increased at 0.05mPa comparedto samples cultured at higher SS or under static conditions. Subjection of cellsto SS higher than 1mPa led to an upregulation of osteogenic differentiation mark-ers [42]. Higher calcium deposition of hMSCs was observed at SS between 0.01mPaand 12mPa compared to samples cultured at SS below 0.01mPa [13].

In the future, ECM growth over time could be included in the simulations aswell. It is known, that SS in a perfused porous structure are highly dependent onits structural properties. Permeability, a factor interrelated with structural factors

98


like porosity, pore size, pore shape etc., has been shown to affect cell proliferation,cellular activity and cellular growth [22]. By implementing the results from μCTmonitoring into the CFD model its effect on porosity and eventually SS could betaken into account. However, using μCT alone, it will still not be possible to assessnon-mineralized ECM growth.

It is important to note, that the results observed are only valid for SF scaffoldswith the reported porosity and permeability, seeded with hMSCs and cultured inthe perfusion bioreactor described. For different scaffold materials, cell types andbioreactor geometries, SS will differ and have to be defined first. CFD simulationswere performed with a simplified scaffold model based on Darcy’s law that is notable to display local differences in scaffold geometry. Differences observed in miner-alized tissue growth between single samples (Fig. 3.3.2) might be explained by localdifferences in scaffold geometry leading to local differences in SS and subsequentdifferences in mineralized tissue growth patterns. Nevertheless, the ROC analysisshowed good agreement of the prediction of mineralized tissue growth.

This study showed a clear dependence of hMSC cell fate on the perfusion velocityapplied. The velocities applied were able to mimic the mechanical environmentduring fracture healing or in healthy bone tissue leading to increased cell proliferationand ECM production (vlow) or mineralized matrix growth (vhigh). Two distinctranges of SS could be defined by CFD showing no overlap of SS between vlow andvhigh. This leads to the assumption that there exist definite SS where hMSCs entercell proliferation or differentiation. Mineralized ECM developed from the edge ofthe scaffolds towards the middle. The optimal SS for mineralized ECM growth isthought to move towards the center of the scaffold due to the filling of the pores onthe edge. By combining the observations from the μCT monitoring with the CFDmodel presented, the mechanical environment may be modeled in the future overthe whole culture period. Together with the optimal SS defined, this study laysthe foundation for a tight control of hMSC cell behavior towards proliferation ordifferentiation in perfusion cultures over the whole culture period.

Conflict of interest statement

The authors have no conflicts of interest to disclose.

Acknowledgements

The research leading to these results has received funding from the Euro-pean Union’s Seventh Framework Programme (FP/2007–2013): FP7-NMP-2010-

99


LARGE-4: BIODESIGN—Rational bioactive materials design for tissue regenera-tion and ERC-2013-StG: REMOTE—Real-time monitoring of load induced remod-eling in tissue-engineered bone. The authors would like to thank Gratianne Vaissonfor the help with cell experiments and Steve Ho for the help with the permeabilitymeasurements of the SF scaffolds.

References

[1] R. L. Duncan and C. H. Turner. Mechanotransduction and the functional-response of bone to mechanical strain. Calcified Tissue Int, 57(5):344–358,1995.





[6] M. J. Jaasma, N. A. Plunkett, F. J. O’Brien. Design and validation of a dynamicflow perfusion bioreactor for use with compliant tissue engineering scaffolds. JBiotechnol, 133(4):490–6, 2008.



[9] V. I. Sikavitsas, G. N. Bancroft, H. L. Holtorf, J. A. Jansen, A. G. Mikos.Mineralized matrix deposition by marrow stromal osteoblasts in 3D perfusionculture increases with increasing fluid shear forces. P Natl Acad Sci USA,100(25):14683–8, 2003.

100


[10] M. E. Gomes, V. I. Sikavitsas, E. Behravesh, R. L. Reis, A. G. Mikos. Effectof flow perfusion on the osteogenic differentiation of bone marrow stromal cellscultured on starch-based three-dimensional scaffolds. J Biomed Mater Res A,67(1):87–95, 2003.

[11] H. L. Holtorf, J. A. Jansen, A. G. Mikos. Flow perfusion culture inducesthe osteoblastic differentiation of marrow stroma cell-scaffold constructs in theabsence of dexamethasone. J Biomed Mater Res A, 72(3):326–34, 2005.

[12] G. M. de Peppo et al. Engineering bone tissue substitutes from human inducedpluripotent stem cells. P Natl Acad Sci USA, 110(21):8680–8685, 2013.


[14] L. S. Gardel, C. Correia-Gomes, L. A. Serra, M. E. Gomes, R. L. Reis. A novelbidirectional continuous perfusion bioreactor for the culture of large-sized bonetissue-engineered constructs. J Biomed Mater Res B, 101(8):1377–1386, 2013.

[15] W. L. Grayson et al. Effects of initial seeding density and fluid perfusion rateon formation of tissue-engineered bone. Tissue Eng Part A, 14(11):1809–1820,2008.


[17] S. Partap, N. A. Plunkett, D. J. Kelly, F. J. O’Brien. Stimulation of osteoblastsusing rest periods during bioreactor culture on collagen-glycosaminoglycan scaf-folds. J Mater Sci Mater Med, 21(8):2325–30, 2010.

[18] R. M. Delaine-Smith, S. Macneil, G. C. Reilly. Matrix production and collagenstructure are enhanced in two types of osteogenic progenitor cells by a simplefluid shear stress stimulus. Eur Cell Mater, 24:162–74, 2012.

[19] K. Kusuzaki et al. Development of bone canaliculi during bone repair. Bone,27(5):655–659, 2000.


[21] D. C. Ding, W. C. Shyu, S. Z. Lin. Mesenchymal stem cells. Cell Transplant,20(1):5–14, 2011.

101


[22] J. Fan, X. Jia, Y. Huang, B. M. Fu, Y. Fan. Greater scaffold permeabilitypromotes growth of osteoblastic cells in a perfused bioreactor. J Tissue EngRegen Med, 2013.


[24] L. Meinel et al. Bone tissue engineering using human mesenchymal stem cells:Effects of scaffold material and medium flow. Ann Biomed Eng, 32(1):112–122,2004.



[27] B. B. Mandal, A. Grinberg, E. S. Gil, B. Panilaitis, D. L. Kaplan. High-strengthsilk protein scaffolds for bone repair. P Natl Acad Sci USA, 109(20):7699–7704,2012.

[28] L. Meinel et al. Silk implants for the healing of critical size bone defects. Bone,37(5):688–98, 2005.






102



[35] E. Zermatten, S. Haussener, M. Schneebeli, A. Steinfeld. Tomography-baseddetermination of permeability and dupuit-forchheimer coefficient of character-istic snow samples. J Glaciol, 57(205):811–816, 2011.

[36] I. Ochoa et al. Permeability evaluation of 45S5 bioglass (R)-based scaffolds forbone tissue engineering. J Biomech, 42(3):257–260, 2009.



[39] F. Maes et al. Computational models for wall shear stress estimation in scaf-folds: A comparative study of two complete geometries. J Biomech, 2012.


[41] E. Zermatten et al. Micro-computed tomography based computational fluiddynamics for the determination of shear stresses in scaffolds within a perfusionbioreactor. Ann Biomed Eng, 42(5):1085–1094, 2014.

[42] B. Porter, R. Zauel, H. Stockman, R. Guldberg, D. Fyhrie. 3-D computationalmodeling of media flow through scaffolds in a perfusion bioreactor. J Biomech,38(3):543–9, 2005.

103

Chapter 4

Implementation of perfusionbioreactor system

105

4.1 Influence of curvature on mineralized matrix

4.1 The influence of curvature on mineralized

matrix formation in 3-D

A model for the investigation of the influence of curvature on three-dimensionalmineralized matrix formation under static and loaded conditions

Jolanda Rita Vetsch1, Ralph Müller1, Sandra Hofmann1,2,3


in preparation

Abstract:

Bone remodelling is the continuous turnover of bone by resorption and formation.It is controlled by interstitial fluid flow sensed by osteocytes. The refilling ofbone resorption sites has been shown to be curvature-driven. In-vitro, curvatureinfluenced tissue growth and cytoskeletal arrangements under static and loadedconditions. Nevertheless, this has only been evaluated for non-mineralized tissuein limited three-dimensional (3-D) volumes. This study aims at investigatingthe influence of three different channel curvatures (S: -2.00mm−1, M: -1.33mm−1,L: -0.67mm−1) on mineralized tissue formation in 3-D under static and loadedconditions. The ingrowth of mineralized tissue into the channels was dependenton curvature and was higher under loading (M and S channels). L channels werenot closed in any group compared to partially (static) or fully (loaded) closed Mand S channels. Interestingly, mineralized tissue morphology was cortical-like instatic samples and trabecular-like in loaded samples. Our results suggest thatthe 3-D in-vitro model presented is not only able to reveal effects of curvature onmineralized tissue formation, but may be used as an in-vitro model for circularcritical size defects in trabecular or cortical bone.

107

Chapter 4 Implementation of perfusion bioreactor system

Keywords:Curvature, micro-computed tomography monitoring, human mesenchymal stemcells, critical size defect, in-vitro model, bone tissue engineering

4.1.1 Introduction

Bone remodelling describes the process of continuous turnover of bone tissue byresorption and bone formation in the adult skeleton, which allows bones to adapttheir structure to changes in long-term mechanical loading [1,2]. The refilling of boneresorption sites with mineralized tissue in-vivo is most likely curvature-driven [3,4].In cortical bone, bone resorption is performed by osteoclasts excavating cylindricaltunnels, which are subsequently refilled with osteoid and a central blood vesselultimately forming a Haversian canal. In trabecular bone, osteoclasts form semi-circular resorption pits that can be seen as hemi-osteons that are later refilled withosteoid produced by osteoblasts [5]. Bone remodelling always leads to transient localchanges of surface geometry of bone tissue [3, 4]. In trabecular bone it has beenshown that average curvature is close to zero, implying a strong control of cellularbehaviour and cell response to changes in tissue geometry [3]. This phenomenon canbe explained by the fact that on curved surfaces, cells are bent and experience tensileforces. Cells aim to reduce these forces by contracting their cytoskeleton whichsubsequently leads to a flattening of the tissue surface. This has been attributed tothe fact that cellular growth is not only controlled by signalling pathways but alsoby physical control mechanisms [6].

The predominant mechanical loading condition in remodelling processes is be-lieved to be the interstitial fluid flow in the lacuno-canalicular network. Interstitialfluid flow is driven by deformations of the bone matrix and sensed by osteocytesthat are also known to orchestrate the process of bone remodelling [7]. In-vitro,mechanical loading of cells with fluid flow using perfusion bioreactors has been per-formed frequently to investigate the effects of fluid flow on bone tissue engineeringcultures [8]. Perfusion has been shown to increase levels of osteogenic markers andmineralized tissue formation of different cell types [9, 10]. Local fluid flow velocityin scaffolds is strongly influenced by scaffold pore size and increases with decreas-ing pore size if the porosity is retained constant [11]. The curvature of a perfectlyspherical pore, defined as the radius−1, is indirect proportional to the pore size andtherefore relates directly to the fluid flow velocity in the pore. Endothelial cellscultured on curved surfaces showed cytoskeletal changes when compared to cellscultured on flat surfaces [12]. The changes observed were significantly enhancedwhen cells were additionally exposed to perfusion, while cells cultured on flat sur-

108


faces remained the same even when perfusion was applied. Taken together, theseobservations lead to the assumption that in-vitro mineralization under loaded con-ditions is also influenced by curvature, but to date, similar experiments using bonecells have not been realized to our knowledge.

Essential work investigating the influence of curvature on non-mineralized tissuegrowth in static cultures was performed by Rumpler et al. [13]. They evaluated thelocal growth rate of non-mineralized tissue formed by osteoblasts on hydroxyapatite(HA) scaffolds of 2mm in height in response to different channel shapes (triangular,square, hexagonal, and circular) and channel sizes (perimeters of 3.14mm, 4.71mm,and 6.28mm). They quantified projected tissue area and visualized the cytoskeletonusing confocal scanning microscopy. Tissue thickness, defined as the layer thick-ness of tissue growing on the channel wall, has been shown to be affected by localcurvature and channel size. Increasing curvature and decreasing channel size led tohigher tissue thickness. The total amount of tissue formed was dependent on thechannel perimeter and was proportional to the local curvature. Other studies havealso shown that the tissue growth rate was highly dependent on curvature [6, 14].Curvature sign was found to be important as well, showing decreasing tissue growthfrom concave (negative curvature) to convex (positive curvature) to flat surfaces(zero curvature) [6, 14–16]. The formation of mineralized tissue on different bioce-ramic materials has been shown to be controlled by the size of cavities introducedto the scaffold’s surface compared to no mineralized tissue formation on the planarsurfaces of the scaffolds [17].

The above mentioned in-vitro studies focused on non-mineralized extracellularmatrix (ECM) formation only [6, 13, 14], or the formation of mineralization wasinitiated by soaking the scaffold material in simulated body fluid which does notreflect cell-mediated mineralization [17]. Despite the fact that 3-D scaffolds havebeen used to investigate the influence of curvature on tissue growth, analyses wereperformed using confocal microscopy limited to a penetration depth of about 500μm[6, 13–15, 18]. To evaluate tissue growth along the whole height of the channels itwould have been necessary to cut the samples into thinner slices by destroying thesamples making them useless for longitudinal monitoring of tissue growth.

In addition to the limited penetration depth of confocal microscopy, measurementsof projected tissue area measurements imply homogeneous tissue growth along theheight of the channel. This seems very unlikely, because the curvature of a channel isdefined by the two principal curvatures in horizontal and vertical direction (see [14]for detailed calculation) and highest tissue growth is therefore expected at half thechannel height. A 3-D computational model of curvature-driven growth was derived

109


from [6] to estimate tissue formation in 3-D. Unfortunately, this model did notcontain any biology, e.g. completely ignoring the influence of growth factors andother chemical signals [15].

These limitations could be overcome by the application of micro-computed tomog-raphy (μCT). μCT is a non-invasive and non-destructive imaging technique [19] thathas been used to investigate the 3-D structure of mineralized biological specimensin-vivo and in-vitro [20–22]. It has been shown that it is possible to longitudi-nally monitor mineralized tissue formation of cells seeded on scaffolds over severalweeks [20].

The aim of this study was to investigate the influence of three different channelcurvatures on mineralized tissue formation in 3-D under static and loaded con-ditions. To quantify the influence of curvature on mineralized tissue formation,μCT online monitoring was applied to calculate static quantitative measures suchas mineralized tissue volume (BV), total volume (TV) or mineralized tissue volumefraction (BV/TV) [23] and dynamic quantitative measures such as mineralized tis-sue formation rate (BFR), or mineralized tissue resorption rate (BRR) [21]. It washypothesized that: (i) mineralized tissue formation is driven by curvature, (ii) min-eralized tissue formation is dependent on the channel depth showing highest value athalf the channel height, and (iii) mechanical loading by perfusion has a synergisticeffect with curvature leading to higher mineralized tissue formation under loadedconditions compared to static conditions.

4.1.2 Methods

Materials

Dulbecco’s Modified Eagle Medium (DMEM), fetal bovine serum (FBS; ordernumber 10270–06), penicillin-streptomycin-fungizone (P/S/F), nonessential aminoacids (NEAA), basic fibroblast growth factor (bFGF), β-glycerolphosphate (βGP),ascorbic acid (AA), and dexamethasone (Dex) were from Gibco (Zug, Switzer-land). Hexafluoroisopropanol (HFIP) was from Fluorochem Ltd (Derbyshire, UnitedKingdom). Lithium Bromide (LiBr) was from Thermo Fisher Scientific (Reinach,Switzerland) and Methanol (MeOH) from Merck (Zug, Switzerland). All other sub-stances were purchased from Sigma (Buchs, Switzerland) and were of analyticalgrade. Silkworm cocoons were kindly supplied by Trudel Silk Inc (Zurich, Switzer-land).

110


Scaffold production

Silk fibroin (SF) scaffolds have been produced as described by [24]. In short, B.mori silkworm cocoons were degummed by boiling twice for 1h in 0.02M Na2CO3

and washed with ultrapure water (UPW). After drying, the silk was dissolved in a9M LiBr solution and dialyzed for 36h against UPW (Slide-A-Lyzer 3.5K MWCO,Thermo Fisher Scientific, Waltham, MA, United States). After freezing at -80◦Cover night, the solution was lyophilized for 5 complete days. The lyophilized silk wasdissolved in HFIP resulting in a 17% (w/v) solution, which was subsequently addedto 2.5g sodium chloride crystals of 300–400μm granule diameter. β-sheet formationwas induced by immersing the silk-salt blocks into 90% MeOH for 30min and sub-sequent drying over night [25]. Before leaching for 2 days in UPW, scaffolds werecut to a height of 3mm using a precision saw (Isomet low speed saw, Buehler, Bluff,IL, United States). The scaffold discs were punched to an outer diameter of 9mmwith the help of a cork borer. Channels and pin holes were punched using biopsypunches of 3mm, 1.5mm and 1mm in diameter (kai medical, Solingen, Germany)with the help of a custom-made metal guide to ensure reproducible positioning ofthe channels within the scaffold volume. The curvatures of the channels were chosenaccording to [13], but with a bigger curvature in L channels leading to relative cur-vatures of +50% (S) and -50% (L) with respect to the curvature of the M channel(Fig. 4.1.1D). Due to the silk’s similar density to the density of the culture medium,scaffolds are not visible in μCT scans. Two pins have been used for scaffold orienta-tion in both bioreactor types and for scaffold fixation in static bioreactors. Scaffoldswere sterilized by steam autoclaving in phosphate buffered saline at 121◦C, 1bar for20min.

Cell Study

Human mesenchymal stem cells (hMSCs; Lonza, Walkersville, MD, United States)were isolated from bone marrow aspirate and characterized as described before [26].P3 hMSCs were expanded for 7 days in expansion medium (DMEM, 10% FBS, 1%P/S/F, 1% NEAA and 1ng/ml bFGF) and resuspended in control medium (DMEM,10% FBS, 1% P/S/F) at a concentration of 5 million cells per 50μl. Scaffolds (N=20)were seeded by pipetting 5 million cells on top of each scaffold. After incubation ina 12-well plate under standard cell culturing conditions (37◦C, 5% CO2) for 90min,half of the scaffolds (N=10) was moved into static bioreactors and supplied with6ml osteogenic medium (control medium, 10mM βGP, 50 μg/ml AA, 100nM Dex)which was completely exchanged 3 times a week. The other half of the scaffoldswas supplied with 1ml control medium and incubated for 24 hours to ensure proper

111


Fig. 4.1.1: Three-dimensional images of mineralized tissue formed in the full volume of the channelswith the corresponding coverage in the void volume for (A) L channels, (B) M channels, and (C) Schannels. (D) Schematic illustration of the scaffold geometry with the three different channel sizesand the two pin holes (left). Radii, perimeters, curvatures and relative curvatures of all channels(middle). Schematic representation of the two volumes (full and void) used for the evaluationof mineralized tissue formation in the channel (right). (E) Coverage of void volume. Scale bar:1mm. *p≤0.05. #p≤0.05. §p≤0.01. # and § indicate statistical significances between differentcurvatures within the same bioreactor type.

cell attachment, before the scaffolds were moved into perfusion bioreactors the nextday. Perfusion bioreactors were provided with 18ml osteogenic medium to avoidair inclusions in the perfusion system. A third of the medium was exchanged 3times a week by triple concentrated osteogenic medium (control medium, 30mMβGP, 150μg/ml AA, 300nM Dex) to keep the concentration of osteogenic factorsconstant. A continuous flow rate of 12ml/min was set at the pump. Moving thescaffolds into the bioreactors was considered as day 0 of the cell culture.

μCT monitoring

The deposition of 3-D mineralized tissue was monitored as described earlier [20]by weekly μCT scans starting at day 7 of the cell culture. All bioreactors werescanned in a μCT scanner at 36μm resolution (SCANCO Medical AG, μCT 40,

112


Brüttisellen, Switzerland). The energy level was set to 45kVp and the intensity to177μA. An integration time of 200ms and 2-fold data averaging was applied. Toreduce noise the reconstructed images were Gaussian filtered with a filter width of1.2 and a support of 1. Mineralized tissue was segmented at a grey-scale value of135 (corresponding to a density of 64.9mg HA per cm3) and unconnected particlessmaller than 20 voxels were removed. 3-D volumes were evaluated for BV/TV asdescribed earlier for bone [19,23].

For final evaluation, all scaffolds exhibiting a BV/TV lower than 1% at day 42 ofthe cell culture were excluded. Due to biological variation, the onset of mineralizedtissue formation was not equal in all bioreactors, which has been shown in previousstudies monitoring mineralized tissue formation in bioreactors [20]. We accountedfor this variation by introducing a threshold value that corresponded to the medianBV/TV value of the full scaffold observed at day 14 of the cell culture (day 14 equaledto the first scan time point where mineralized tissue formation was observed). Theday when the BV/TV of a single bioreactor exceeded the threshold value was definedas time point 0 (t0). After passing the threshold each bioreactor was maintained inculture for additional 4 weeks, leading to a total of 5 evaluated scan time points foreach bioreactor (t0, t1, t2, t3, t4). All evaluations were based on these time points.

Mineralized tissue formation: full scaffold

Mineralized tissue formation in the full scaffold was determined by fitting a cylin-drical mask to the outer scaffold borders. The volume of the fixation pins wasdetermined from day 7 scans and subsequently subtracted from BV for later timepoints. Site-specific variations in vertical distribution of mineralized tissue in thefull scaffold were evaluated by regionalizing the full volume into 10 regions and 9overlapping regions (Fig. 4.1.2D). For each region BV/TV was quantified and amorphometrical profile was generated.

Mineralized tissue formation: channels

Mineralized tissue formation in the three channels was evaluated based on two dif-ferent volumes of interest: (i) the channel void volume and (ii) the channel volumeincluding an adjacent ring volume comprising of the SF scaffold (Fig. 4.1.1D). Asfrom now these volumes will be referred to as ’void volume’ and ’full volume’. Thediameter of the void volume was set to 90% of the real channel diameter to excludepotential effects of the scaffold-channel interface (S: 0.9mm; M: 1.35mm; L: 2.7mm).The diameter of the full volume was set to the real channel diameter plus 1mm (S:2mm; M: 2.5mm; L: 4mm). This volume was chosen in order to investigate the

113


Fig. 4.1.2: Three-dimensional images of mineralized tissue formed in the full scaffold under: (A)static conditions and (B) loaded conditions. (C) Mineralized tissue volume fraction (BV/TV) ofthe full scaffold. (D) Schematic illustration of regionalization of the scaffold volume into 10 regionsand 9 overlapping regions (top) and the corresponding morphometrical profile. Scale bar: 1mm.*p≤0.05. **p≤0.01.

influence of curvature on mineralized tissue formation in the SF scaffold close bythe channel. The mineralized tissue in the void volume was projected in top-bottomdirection creating a pseudo-radiograph to calculate the coverage in analogy to [27].The coverage was calculated to quantify the ingrowth of mineralized tissue into thechannel. The effect of curvature on dynamic mineralized tissue formation param-eters was investigated using a technique that has been used to quantify dynamicbone morphometric parameters in-vivo [21, 28]. Briefly, dynamic mineralized tis-sue formation parameters were determined by registering μCT scans of t2 and t4.Based on the registered μCT scans, sites that were present at t2 only have beenconsidered as removed mineralized tissue (coloured in blue), sites present at t4 onlyhave been considered as newly formed mineralized tissue (coloured in orange) andsites present at both t2 and t4 have been considered as constant mineralized tissue(coloured in grey; Fig. 4.1.3A-C, G-I). The registered μCT scans were further eval-uated for the calculation of dynamic mineralized tissue morphometric parametersnamely mineralized tissue formation rate (BFR), mineralizing surface (MS), andmineral apposition rate (MAR; see [21] for detailed description of parameter calcu-lation). BFR is defined as the formed BV per existing BV per day in percentage(%/day), MS is defined as the percentage of formed mineralized BS per total existingBS (%), and MAR is defined as the mean thickness of formed BV per day in microns

114


(μm/day) in accordance with [21]. Dynamic parameters have been determined forthe void volume (Fig. 4.1.3D-F) as well as for the full volume (Fig. 4.1.3J-L).

Fig. 4.1.3: Three-dimensional images of registered scans of time point 2 and time point 4 showingnewly formed (orange), constant (grey), and resorbed (blue) mineralized tissue for L channels (A,J), M channels (B, K), and S channels (C, L) in the full volume and the void volume, respectively.Dynamic mineralized tissue morphometric parameters for the full volume and the void volume aredepicted as (D, J) mineralized tissue formation rate (BFR), (E, K) mineralizing surface (MS), and(F, L) mineral apposition rate (MAR). Scale bar: 1mm. *p≤0.05. **p≤0.01

115


Histology

At the end of the study scaffolds were fixed in 10% (v/v) neutral buffered formalinover night at 4◦C. After embedding in paraffin scaffolds were cut to a thickness of10μm into either horizontal cross-sections through the top, middle and bottom ofthe scaffold or into vertical cross-sections through all channel sizes. To visualize cellnuclei and ECM, Haematoxylin&Eosin (H&E) staining was performed. Sirius Redstaining was performed to visualize collagen.

Statistics

All statistical evaluations were performed using IBM SPSS Statistics 20 (SPSS Inc.,Chicago IL, United States). Comparisons of more than two means were done byrepeated-measures analysis of variance (ANOVA) or ANOVA followed by post-hoc testing with Bonferroni corrected significance levels. Data is presented asmeans±standard deviation and was considered statistically significant at p≤0.05and highly statistically significant at p≤0.01. Figures are showing upper mediansamples.

4.1.3 Results

Mineralized tissue formation: full scaffold

3-D images showed substantial mineralized tissue formation in the static(Fig. 4.1.2A) and the loaded group (Fig. 4.1.2B). L channels were clearly visiblein both groups. M channels of the static group were clearly visible. S channels andthe M channel of the loaded group were less visible especially at late time points(t3, t4) due to the channels filling up with mineralized tissue over time. BV/TVof the static group was significantly higher than BV/TV of the loaded group forall time points (p≤0.01). BV/TV of the static group increased from 5.70%±2.09%at t0 to 21.04%±1.65% at t4, whereas BV/TV of the loaded group increased from1.94%±0.40 at t0 up to 6.97%±2.47% at t4 (Fig. 4.1.2C). Mineralized tissue formedin the static group was mostly located at the top and on the edge of the scaffold,whereas mineralized tissue formed in the loaded group was distributed more homo-geneously throughout the full scaffold volume. The morphometrical profile showeda regional dependence of BV/TV (p≤0.01; Fig. 4.1.2D). The static group showedan increased BV/TV value in the upper regions (regions 3 to 10) compared to theloaded group. The mineralized tissue of the static group was unevenly distributedalong the scaffold height. More mineralized tissue was found in the upper regions(regions 3 to 7), where a higher BV/TV value was observed compared to at least

116


3 other regions (p≤0.05; significances not indicated in figure). In the loaded groupBV/TV was identical among all regions. These results confirm that the distributionof mineralized tissue in the scaffold is highly dependent on the loading conditionapplied.

Mineralized tissue formation: channels

3-D images of mineralized tissue in the full volume (=channel volume plus an adja-cent ring volume comprising of the SF scaffold) showed that all channels in the staticgroup and the L channels in the loaded group were not completely filled with miner-alized tissue (Fig. 4.1.1A-C). In the loaded group the M channels were almost closed(Fig. 4.1.1B). The S channels were completely closed in both groups (Fig. 4.1.1C).Even more pronounced as for the full scaffold, the mineralized tissue in the staticgroup was mostly located at the top of the scaffold unlike hypothesized. Mineralizedtissue in the loaded group was distributed throughout the whole scaffold height. Thecoverage, representing tissue ingrowth into the void volume (=void volume of thechannel with the diameter set to 90% of the real channel diameter), confirmed theresults observed in 3-D images (Fig. 4.1.1A-C). Curvature had a significant influenceon coverage (p≤0.01; Fig. 4.1.1E). The coverage in the loaded group was higher in L(p≤0.01) and M (p≤0.05) channels compared to the coverage in L and M channels ofthe static group. The static group showed a higher coverage in S channels comparedto L channels (p≤0.05), whereas the loaded group showed a higher coverage in Sand M channels compared to L channels (p≤0.01).

3-D images of registered scans from t2 and t4 revealed newly formed mineralizedtissue (orange), constant mineralized tissue (grey) and—as expected—very littleremoved mineralized tissue (blue; Fig. 4.1.3A-C, G-I). In the full volume, BFRwas dependent on the loading condition applied (p≤0.01), but there were no differ-ences observed between the different curvatures (Fig. 4.1.3D). The loading condition(p≤0.01) and the curvature (p≤0.05) influenced the MS, but no differences were ob-served among the different curvatures or loading conditions (Fig. 4.1.3E). MAR wasinfluenced by the loading condition (p≤0.01) and was lower for L channels in theloaded group (p≤0.05; Fig. 4.1.3F). In the void volume, no significant differenceswere observed for BFR and MAR (Fig. 4.1.3J, L). MS was influenced by the loadingcondition (p≤0.05), but no differences were observed for curvature (Fig. 4.1.3K).

Histology

Horizontal scaffold cross-sections stained with H&E revealed almost no ECM in-growth in L channels for both loading conditions neither at the top nor the bottom

117


of the scaffold (Fig. 4.1.4A, D, G, I). In the static group ECM ingrowth was ob-served in M and S channels. M channels were almost closed at the top and lessclosed at the bottom of the scaffold (Fig. 4.1.4B, E). S channels were completelyclosed at the top, but not completely at the bottom of the scaffold (Fig. 4.1.4C,F). In the loaded group, M and S channels were completely closed independent ofthe location in the scaffold (Fig. 4.1.4H, I, K, L). The more uniform ECM distri-bution along the height of the scaffold in the loaded group was in agreement withthe more homogeneous distribution of mineralized tissue observed with μCT. Simi-lar patterns were observed for horizontal scaffold cross-sections stained with SiriusRed (Fig. 4.1.5). Collagen was present in the static and the loaded group. Verticalscaffold cross-sections through S channels confirmed the results observed for hori-zontal cross-sections (Fig. 4.1.6). In the static group, S channels were closed in theupper half of the scaffold only (Fig. 4.1.6A), whereas in the loaded group S channelswere completely filled (Fig. 4.1.6B). Again, collagen staining was positive for bothconditions following the same patterns as observed for H&E staining (Fig. 4.1.6C,D).

4.1.4 Discussion

The process of bone remodelling characterized by resorption and formation con-stantly leads to transient changes of bone surface geometry [1,2]. It was shown thatthe average curvature of trabecular bone is close to zero indicating an intimate cor-relation between geometrical changes and cellular behaviour in-vivo [3]. The boneremodelling process is thought to be mechanically controlled by interstitial fluid flowsensed by osteocytes [7]. Different experimental studies have shown that curvatureinfluences tissue growth under static and loaded conditions in-vitro [6, 12–14, 18].Projected tissue area, tissue thickness and tissue growth rate was highly dependenton local curvature under static conditions [13]. Changes in cytoskeletal arrangementwere more pronounced when cells were cultured on curved surfaces under loaded con-ditions [12]. Nevertheless, these studies focused on non-mineralized ECM formationonly and evaluations of tissue growth were performed in limited 3-D volumes. Thestudy presented herein investigated the influence of three different channel curva-tures on 3-D mineralized tissue formation. Scaffolds with three channels correspond-ing to the three different curvatures (S: -2.00mm−1, M: -1.33mm−1, L: -0.67mm−1)have been produced (Fig. 4.1.1D). 3-D mineralized tissue formation was investigatedusing μCT monitoring in the full scaffold volume (including all channels; Fig. 4.1.2)and two different sub-volumes: (i) the ’void volume’: representing the void volumeof the channel, and the ’full volume’: representing the channel volume including an

118


Fig. 4.1.4: Horizontal scaffold cross-sections stained with Haematoxylin&Eosin. The dashed linedcircles represent the real diameter of the channels (L: 3mm, M: 1.5mm, S: 1mm). (A-C) L, M andS channels of scaffolds cultured under static conditions at the top of the scaffold. (D-F) L, M andS channels of scaffolds cultured under static conditions at the bottom of the scaffold. (G-I) L, Mand S channels of scaffolds cultured under loaded conditions at the top of the scaffold. (J-L) L,M and S channels of of scaffolds cultured under loaded conditions at the bottom of the scaffold.Scale bar: 500μm.

adjacent ring volume comprising of the SF scaffold (Fig. 4.1.1, Fig. 4.1.3).The BV/TV in the full scaffold volume was always higher in the static group com-

pared to the loaded group (Fig. 4.1.2C). This may be attributed to the mechaicalloading condition applied being not optimal for increased mineralized tissue forma-tion. Some studies have shown that mechanical loading by perfusion may increasemineralized matrix formation of cells in culture, but other studies reported cellproliferation rather than differentiation or even apoptosis depending on the flow

119


Fig. 4.1.5: Horizontal scaffold cross-sections stained with Sirius Red. The dashed lined circlesrepresent the real diameter of the channels (L: 3mm, M: 1.5mm, S: 1mm). (A-C) L, M and Schannels of scaffolds cultured under static conditions at the top of the scaffold. (D-F) L, M andS channels of scaffolds cultured under static conditions at the bottom of the scaffold. (G-I) L, Mand S channels of scaffolds cultured under loaded conditions at the top of the scaffold. (J-L) L,M and S channels of of scaffolds cultured under loaded conditions at the bottom of the scaffold.Scale bar: 500μm.

velocity applied [29, 30]. Lower BV/TV values observed for the full scaffold in theloaded group indicate that the perfusion culture conditions chosen might have beennot optimal to enhance mineralized tissue formation but rather have induced cellproliferation. To investigate the effects of loading on the morphology and spatial dis-tribution of the mineralized tissue formed, the total amount of mineralized tissue ofis of minor relevance in this study. In static cultures cells tend to concentrate on theouter scaffold surface due to a poor nutrient and waste exchange to the centre of the

120


Fig. 4.1.6: Vertical scaffold cross-sections stained with Haematoxylin&Eosin (A, B) and Sirius Red(C, D). The dashed lines represent the width of the S channel (1mm). Scale bar: 500μm.

scaffold [31]. This is still a major issue in BTE and several studies tried so solve thisproblem using dynamic bioreactors or the engineering of vascularisation [32,33]. Theapplication of perfusion improved the mineralized tissue distribution in the loadedgroup vastly. In the static group the mineralized tissue was located at the top ofthe scaffold and on the edge of the scaffold (Fig. 4.1.2A, D). It may be hypothesizedthat the cells were not able to migrate into the scaffold volume after seeding them ontop of the scaffold or that the cells could have died in the middle of the scaffold dueto a lack of nutrient and waste exchange [34]. These assumptions were disprovenby histological data showing cells throughout the whole scaffold volume (Fig. 4.1.5,Fig. 4.1.6). The enhanced tissue distribution in the loaded group might be due toimproved nutrient and waste exchange in the scaffold volume, giving rise to a moreadvantageous environment for the cells to produce mineralized tissue.

The mineralized tissue formation in the channels was influenced by curvature(Fig. 4.1.1A-C, E). The ingrowth of mineralized tissue into the void volume of thechannel, represented by coverage, was directly dependent on curvature. As hy-pothesized, mineralized tissue formation was increased with increasing curvature inanalogy to higher projected tissue area with increasing curvature reported for non-mineralized ECM before [6,13,14,18]. These results have been confirmed by histol-ogy. H&E staining showed that cells were distributed throughout the whole scaffoldvolume under static and dynamic loading conditions. The closing of the channelswith ECM was in agreement with the mineralized tissue formation observed by μCT.Sirius Red staining verified the presence of collagen under both culture conditionsin the whole scaffold volume following the same pattern as observed for ECM with

121


H&E staining. In-vivo, the formation of mineralized tissue is a sequential process,initiated by the formation of a 3-D collagen framework, which is the major com-ponent of the bone matrix. Subsequently mineralized tissue is deposited on theframework [35, 36]. Curvature is thought to increase ECM and collagen productionat first, which subsequently leads to increased deposition of mineralized tissue. Thepresence of collagen could therefore be evidence for future mineralized tissue forma-tion locations. This might be an indication that curvature influences ECM formationdirectly and mineralized tissue formation indirectly via increased ECM formation.A limitation of this study is that collagen cannot be visualized with μCT due to itslow density and could only be determined at the last time point of this study byhistology, which is a destructive technique. Nevertheless, the evaluation of severalsequential μCT scans in combination with computational simulations could lead toa predictive model revealing potential sites for newly mineralized tissue formation.

The mineralized tissue formation was expected to be different depending on thedepth of the channel. According to simulations of curvature-driven growth in 3-Dcylindrical channels the highest mineralized tissue growth was predicted at half thechannel height and the final shape of the tissue formed should resemble a catenoid(a surface that is generated by rotating a U-shaped curve formed by a rope about itshorizontal axis, here the inner wall of the channel) [14,15]. The shape of the miner-alized tissue in the void volume of all channels was inhomogeneous and did not showa catenoid pattern neither under static nor under loaded conditions (Fig. 4.1.1A-C).This implies that the mineralized tissue growth inside a cylindrical channel doesnot follow the simulation [15]. Furthermore, the simulations were performed understatic conditions only and are therefore not able to explain mineralized tissue for-mation in loaded cultures. Under static conditions, the positive effects of curvaturedo not lead to increased mineralized tissue growth within the cylindrical channels.The application of perfusion loading though was able to improve mineralized tissuegrowth and spatial distribution within the channels.

Culture conditions showed distinct influences on tissue ingrowth into the channels.L channels were not closed in the static and the loaded group. Similar patterns wereobserved in M and S channels, but with higher tissue ingrowth into the void volumeas represented by increasing coverage with decreasing channel size (Fig. 4.1.1E). Mand S channels of the loaded group showed mineralized tissue formation throughthe full channel diameter despite there was no scaffold material present. It wasexpected that the loading would enhance mineralized tissue deposition having asynergistic effect with higher curvature, but this could not be shown. Endothelialcells cultured on curved surfaces showed cytoskeletal changes which were additionally

122


enhanced by the application of fluid flow [12]. The observed cytoskeletal changesin endothelial cells however, might not be beneficial to increase the mineralizedtissue formation by hMSCs. This cannot be confirmed by this experiment becausecytoskeletal changes were not examined. The investigation of cytoskeletal changesrequires destructive techniques and would not have been feasible to perform duringthe experiment presented. Investigating cytoskeletal changes might be consideredfor future experiments to study the influence of loading and curvature on mineralizedmatrix production with respect to potential cytoskeletal changes of hMSCs.

As described for the full scaffold volume, culture conditions influenced the mor-phology of the mineralized tissue in and close to the channels (Fig. 4.1.1A-C). Understatic conditions, the mineralized tissue was more solid and located at the edge ofthe void volume resembling cortical bone. In the loaded group, the mineralizedtissue was more porous resembling trabecular bone and the mineralized tissue wasdistributed more homogeneously over the void volume. Like for the full scaffoldvolume the difference in tissue distribution is most probably attributable to thedifferences in nutrient and waste transport.

Dynamic mineralized tissue morphometric parameters have been little influencedby curvature or culture condition (Fig. 4.1.3D-F, J-L). Some non-significant trendscould be observed for the full volume though: (i) BFR was slightly higher in per-fusion bioreactors (Fig. 4.1.3D), (ii) MS seemed to be increased with decreasingpore size (Fig. 4.1.3E), and (iii) MAR tended to be higher in static bioreactors(Fig. 4.1.3F). It may be assumed that the curvature does not only affect tissueingrowth into the void volume of the channel, but also affects mineralized tissueformation in the SF scaffold in close proximity to the channel. This is underlinedby the fact that no trends could be observed for the void volume (Fig. 4.1.3J-L).For non-mineralized tissue it was shown that ECM formation rate is curvature de-pendent [6, 14], but BFR did not follow this behaviour. The higher tissue ingrowthwith increasing curvature can therefore not be explained by higher BFR, but is mostprobably attributable to an increased MS. Dynamic morphometric parameters havebeen observed in-vivo before [21,28]. In-vivo mechanical loading has been shown toincrease BFR and MS [21]. A similar effect was anticipated for the study performedto increase the rate of tissue repair rate. It is assumed that the in-vitro model de-scribed is no able to reproduce the effects observed in-vivo due to a lack of othercell types, growth factors, or optimal loading conditions.

The curvature might not be represented correctly in the pores due to the distortionof the scaffolds, which is a major limitation of this study. The distortion of thescaffolds might be attributable to the fact that the scaffolds could have been under

123


compression while punching using the custom-made metal guide. Also, the fixationof the scaffolds in the bioreactors could have influenced their shape. It needs tobe assumed that the channels in the scaffolds were not always perfectly round,which was confirmed by histology images (Fig. 4.1.5, Fig. 4.1.6). The curvatureinside a channel might therefore not be the same for all the cells in this channel.This could explain inhomogeneous ingrowth patterns of mineralized tissue observed(Fig. 4.1.1A-C).

The main purpose of this paper was to investigate the influence of curvatureon mineralized tissue formation in 3-D. Another interesting aspect of this studythough, is the resemblance of the channels to circular bone defects. With the modelpresented the size of a critical defect could be imitated. By classical definition, acritical size defect is the smallest size tissue defect that will not heal completelyover the lifetime of an animal [37]. The channels of our scaffold closely resemblecranial defects and the ingrowth of mineralized tissue into the channels followed thesame pattern as observed for cranial defect healing in-vivo [38]. Critical size cranialdefects have been defined for mice as 5mm or rats as 8mm round defects [39]. Adultcranial defects in human do not heal spontaneously thus, any clinically significantdefect is a critical defect [40]. From a mechanical point of view, the static conditionrepresents a cranial defect closer compared to the loaded condition, because skullbones normally experience minimal stresses or strains [41] . From a structural pointof view, static conditions could be used to simulate cortical bone defects and loadedconditions for trabecular bone defects. The experimental design might even allowfor scaffold material testing. The channels could be filled with different materials toinvestigate their effect on bone regeneration in a simulated defect.

4.1.5 Conclusion

In conclusion, the results of this study demonstrate the influence of curvature onmineralized tissue formation in 3-D. Mineralized tissue formation was dependenton curvature which was represented by an increased coverage of S and M channelscompared to L channels. The amount of mineralized matrix formation was notdependent on the depth of the channel and was not increased by perfusion biore-actor culture. Nevertheless, mineralized tissue morphology was influenced by theculture condition (static, loaded), with more solid, cortical-like morphology understatic conditions or more porous, trabecular-like morphology under loaded condi-tions. Taken together, with the in-vitro model described, cylindrical bone defects oftrabecular of cortical bone with or without scaffold material could be simulated in-vitro prior to conducting in-vivo experiments, possibly leading to a reduced number

124


of animal experiments. Additionally, the knowledge gained about curvature-drivenmineralized tissue formation in 3-D could be applied to specifically design scaffoldmorphologies for bone tissue engineering or in-vivo bone regeneration.

4.1.6 Competing interests

We have no competing interests.

4.1.7 Authors’ contributions

JRV carried out sequence alignments, designed and coordinated the study, carriedout the lab work, performed data analysis, carried out the statistical analyses, anddrafted the manuscript. RM helped to carry out sequence alignments, participatedin the design of the study and revised the manuscript critically. SH participatedin carrying out sequence alignments, helped to design the study and revised themanuscript thoroughly and critically. All authors gave final approval for publication.

4.1.8 Funding

This project has been funded by the European Union’s Seventh FrameworkProgramme (FP/2007–2013): FP7-NMP-2010-LARGE-4: BIODESIGN—Rationalbioactive materials design for tissue regeneration.

References



[3] H. Jinnai et al. Surface curvatures of trabecular bone microarchitecture. Bone,30(1):191–194, 2002.

[4] A. G. Robling, A. B. Castillo, C. H. Turner. Biomechanical and molecularregulation of bone remodeling. Annu Rev Biomed Eng, 8:455–98, 2006.

[5] A. M. Parfitt. Osteonal and hemi-osteonal remodeling: the spatial and temporalframework for signal traffic in adult human bone. J Cell Biochem, 55(3):273–86,1994.

125


[6] C. M. Bidan et al. How linear tension converts to curvature: Geometric controlof bone tissue growth. Plos One, 7(5):e36336, 2012.






[12] M. D. Frame and I. H. Sarelius. Flow-induced cytoskeletal changes in endothe-lial cells growing on curved surfaces. Microcirculation, 7(6 Pt 1):419–27, 2000.

[13] M. Rumpler, A. Woesz, J. W. Dunlop, J. T. van Dongen, P. Fratzl. The effect ofgeometry on three-dimensional tissue growth. J R Soc Interface, 5(27):1173–80,2008.

[14] C. M. Bidan et al. Geometry as a factor for tissue growth: Towards shapeoptimization of tissue engineering scaffolds. Adv Healthc Mater, 2(1):186–194,2013.

[15] C. M. Bidan, F. M. Wang, J. W. Dunlop. A three-dimensional model fortissue deposition on complex surfaces. Comput Methods Biomech Biomed Engin,16(10):1056–70, 2013.

[16] E. Gamsjager, C. M. Bidan, F. D. Fischer, P. Fratzl, J. W. Dunlop. Modellingthe role of surface stress on the kinetics of tissue growth in confined geometries.Acta Biomater, 9(3):5531–43, 2013.

[17] M. Bianchi et al. Substrate geometry directs the in vitro mineralization ofcalcium phosphate ceramics. Acta Biomaterialia, 10(2):661–669, 2014.

126


[18] K. P. Kommareddy et al. Two stages in three-dimensional in vitro growth oftissue generated by osteoblastlike cells. Biointerphases, 5(2):45–52, 2010.



[21] F. A. Schulte, F. M. Lambers, G. Kuhn, R. Muller. In vivo micro-computed to-mography allows direct three-dimensional quantification of both bone formationand bone resorption parameters using time-lapsed imaging. Bone, 48(3):433–42,2011.

[22] M. Stauber and R. Muller. Micro-computed tomography: a method for thenon-destructive evaluation of the three-dimensional structure of biological spec-imens. Methods Mol Biol, 455:273–92, 2008.





[27] M. P. Lutolf et al. Repair of bone defects using synthetic mimetics of collagenousextracellular matrices. Nat Biotechnol, 21(5):513–8, 2003.

[28] F. M. Lambers, G. Kuhn, F. A. Schulte, K. Koch, R. Muller. Longitudinalassessment of in vivo bone dynamics in a mouse tail model of postmenopausalosteoporosis. Calcif Tissue Int, 90(2):108–19, 2012.

127



[30] S. Partap, N. A. Plunkett, D. J. Kelly, F. J. O’Brien. Stimulation of osteoblastsusing rest periods during bioreactor culture on collagen-glycosaminoglycan scaf-folds. J Mater Sci Mater Med, 21(8):2325–30, 2010.


[32] M. I. Santos and R. L. Reis. Vascularization in bone tissue engineering: physiol-ogy, current strategies, major hurdles and future challenges. Macromol Biosci,10(1):12–27, 2010.

[33] R. I. Abousleiman and V. I. Sikavitsas. Bioreactors for tissues of the muscu-loskeletal system. Adv Exp Med Biol, 585:243–59, 2006.


[35] E. J. Mackie. Osteoblasts: novel roles in orchestration of skeletal architecture.Int J Biochem Cell Biol, 35(9):1301–5, 2003.

[36] S. Weiner and P. M. Dove. An overview of biomineralization processes and theproblem of the vital effect. Biomineralization, 54:1–29, 2003.

[37] P. P. Spicer et al. Evaluation of bone regeneration using the rat critical sizecalvarial defect. Nat Protoc, 7(10):1918–29, 2012.

[38] J. U. Umoh et al. In vivo micro-CT analysis of bone remodeling in a rat calvarialdefect model. Phys Med Biol, 54(7):2147–2161, 2009.

[39] G. M. Cooper et al. Testing the critical size in calvarial bone defects: Revisitingthe concept of a critical-size defect. Plast Reconstr Surg, 125(6):1685–1692,2010.

[40] C. C. Teng, L. G. Shapiro, R. A. Hopper, J. V. Halen. Pediatric cranial defectsurface analysis for craniosynostosis postoperation ct images. 2008 Ieee Inter-

128


national Symposium on Biomedical Imaging: From Nano to Macro, Vols 1-4,pages 620–623, 2008.

[41] D. R. Carter. Mechanical loading histories and cortical bone remodeling. CalcifTissue Int, 36:S19–S24, 1984.

129

Chapter 5

Synthesis

131

Background

In orthopaedics, conventional strategies for bone replacement are patient-derived(autologous) or donor-derived (allogeneic, xenogeneic) bone grafts. At the presentday, the replacement of large bone defects referable to trauma or congenital diseasesis still a major issue due to several drawbacks like limited supply, post-operativecomplications or donor site morbidity [1]. The research field of bone tissue engineer-ing (BTE) tries to overcome these drawbacks. BTE is an evolving field of researchaiming at creating functional constructs for bone replacement grown by cells seededon a scaffold material [2].

Bioreactors have been used for more than 25 years to generate functional cell-matrix constructs in-vitro. Bioreactors allow a high degree of reproducibility, controland automation [3]. Dynamic bioreactors have been designed to overcome the draw-backs of static in-vitro cultures, such as poor nutrient exchange and waste removalin the middle of the scaffold leading the tissue concentrated to the outer scaffoldborders and to cell death in the centre of the constructs [4]. Various dynamic biore-actor designs have been proposed to overcome these limitations and to additionallyapply mechanical loading in BTE cultures.

In-vivo, mechanical loading plays an important role in bone remodelling and frac-ture healing. Healthy bone is constantly undergoing remodelling processes of boneformation and bone resorption adapting its histological structure to changes in long-term mechanical loading [5]. These processes are orchestrated by osteocytes thatare able to sense their mechanical environment [6]. Osteocytes are sitting withinthe lacuno-canalicular system, which is filled with interstitial fluid. Driven by bonematrix deformations the interstitial fluid flow creates shear stresses (SS), which aresensed by osteocytes most likely due to conformational changes in different cellu-lar structures [6, 7]. Cells within repair tissue are exposed to flow induced SS aswell [8]. The loading regime present has been shown to influence tissue developmentof bone regeneration constructs [9]. Similar to osteocytes in healthy bone, progenitorcells like human mesenchymal stem cells (hMSCs) play a major role during fracturerepair. The mechanosensitivity of hMSCs was proven by showing osteogenic dif-ferentiation of hMSCs when exposed to SS [10, 11].The ease of harvesting hMSCsand their large in-vitro proliferation potential makes them a suitable cell source forclinical applications [8]. The principle of mechanical loading in bone is consideredto be mimicked by perfusion bioreactors the closest [12] and could therefore be acrucial means to improve the structure and functionality of in-vitro engineered boneconstructs [2].

In-vitro experiments using perfusion bioreactors have shown to enhance osteogenic

133

Chapter 5 Synthesis

differentiation, increase levels of osteogenic markers and enhance mineralized matrixdeposition of various cell types [4,10,13]. Despite these beneficial features of perfu-sion bioreactors, several studies reported increased cell proliferation rather than celldifferentiation or even apoptosis depending on the perfusion velocities applied [4,10]This suggests that various influencing parameters of perfusion bioreactor studiesin BTE are not completely understood, yet. To understand biological processes inBTE cultures it is necessary to determine the influence of environmental cues on theexperimental outcome.

Micro-computed tomography (μCT) has the potential to add high value to BTEcultures allowing a quantitative and non-invasive investigation of mineralized tissueformation or scaffold geometries in three dimensions (3-D) [14]. The combination ofcomputational simulations with μCT has shown to be a powerful tool providing de-tailed and accurate computations of various aspects of dynamic BTE cultures [15,16].In literature, computational approaches have been described for the simulations ofdynamic BTE determining tissue growth [17], solute concentration [18], or mechan-ical environment [19].

Taken this into consideration, the final aim of this thesis was to develop a perfusionbioreactor system for BTE applications with the possibility for μCT application.First, the original design of our in-house designed perfusion bioreactor was adaptedbased on computational fluid dynamics (CFD) simulation results aiming at providinga homogeneous mechanical environment. Second, cell culture conditions for BTEcultures, namely culture medium supplementation and mechanical loading regime,were evaluated and optimized for the perfusion bioreactor design developed. Third,the optimized perfusion bioreactor system was applied to explain the influence ofscaffold geometry on mineralized tissue formation by the example of curvature. Theinsights gained from this thesis may lead to an improved understanding of causalrelations between different influencing parameters like mechanical environment orculture medium composition and experimental outcomes of dynamic BTE cultures.

Main findings and implications

The first major achievement was the optimization of an in-house designed perfu-sion bioreactor to provide a homogeneous mechanical environment in the bioreactorand the scaffold (chapter 3.1). It was shown that velocity fields inside the biore-actor and the scaffold were inhomogeneous especially at higher flow rates. Theseinhomogeneities were attributable to the original design of the perfusion bioreactor.Three different approaches have been simulated to improve the homogeneity of the

134

velocity fields within the bioreactor and the scaffold: (i) increasing the height of thebioreactor, (ii) increasing the diameter of the inlet and outlet, or (iii) the insertionof flow conditioners. All three approaches led to homogeneous velocity fields in thebioreactor and the scaffold. Nevertheless, increasing the height of the bioreactor wasnot feasible for μCT applications and the insertion of flow conditioners led to air in-clusion within the bioreactor. Therefore, the original bioreactor design was adaptedby increasing the diameter of the inlet and outlet. A high-resolution μCT-basedfinite element simulation of the mechanical environment of a silk fibroin (SF) scaf-fold within the perfusion bioreactor design developed was performed and publishedelsewhere [20]. The next achievement was the particular investigation of the effectof culture medium supplementation with fetal bovine serum (FBS) on mineralizedtissue formation on SF scaffolds (chapter 3.2). It is known that there are differencesin chemical composition of FBS between different batches or brands and these dif-ferences have also been shown to affect experimental outcome [21]. The mineralizedtissue formation observed was highly dependent on the FBS type used. We couldshow that some FBS types are able to induce spontaneous mineralized tissue for-mation on acellular SF scaffolds. Spontaneous mineralization occurred on acellularscaffolds only and the formation of mineralized tissue in cell-seeded scaffolds wascell-mediated. The general aim of BTE engineering is to build functional constructsin-vitro. Morphology and mechanical properties are just two of the plethora of prop-erties defining functional tissue-engineered bone constructs. It might therefore notbe suitable to produce as much mineralized tissue on acellular scaffolds, because itmight lack the correct mechanical properties for bone regeneration, possibly leadingto stress shielding, or blood vessels or cells would not able to invade the tissue dueto a too high density of the mineralized matrix [22]. Regarding clinical applicationshuman serum would be an interesting alternative but due to its high variability be-tween different donors and limited availability it is not an reasonable alternative toanimal serum [23] and it is therefore abundant to look for serum-free alternatives.

In chapter 3.3 the influence of the mechanical loading regime on cellular behaviourof hMSCs was presented. It was observed that the mineralized tissue formed byhMSCs cultured on SF scaffolds is dependent on the flow rate applied. Interestingly,low flow rates induced proliferation of hMSCs, whereas high flow rates inducedmineralized tissue formation by hMSCs. These results show that there are differentranges of mechanical loading leading to either cell proliferation or differentiation.In addition, computational modelling was performed to estimate SS acting on thecells in culture. Two ranges of SS were defined to induce either proliferation ordifferentiation of hMSCs. The SS defined might be used for future perfusion studies

135

Chapter 5 Synthesis

to guide the cellular behaviour of hMSCs depending on the aim of the study.

The perfusion bioreactor system developed was ultimately applied with the opti-mized culture conditions in an in-vitro study to investigate the influence of curvatureunder static and loaded conditions on 3-D mineralized tissue formation (chapter 4).This study was motivated by the fact that in-vivo, the refilling of bone resorptionsites has been shown to be dependent on geometrical changes in the bone matrix.The curvature of trabecular bone is closely conserved to zero implying a strong con-trol of cellular behaviour by changes in tissue geometry [24]. SF scaffold with chan-nels of three different curvatures were produced. It was shown that 3-D mineralizedtissue formation was directly dependent on curvature. The ingrowth of mineralizedtissue into the channels was increased with increasing curvature and was addition-ally enhanced by mechanical loading. Interestingly, the loading condition affectedthe morphology of the mineralized tissue formed showing a solid, cortical-like struc-ture under static conditions and a porous, trabecular-like structure under loadedconditions. With the study performed it was shown that the perfusion bioreactorsystem developed was successfully applied to investigate the influence of curvatureand mechanical loading condition on 3-D mineralized tissue formation of hMSCs onSF scaffolds.

Limitations and future outlook

The thesis presented displays several limitations that need to be considered. Ageneral limitation of the optimization of the perfusion bioreactor design describedin chapter 3.1 is the simplified CFD model. The real scaffold geometry was notincluded and was simplified according to Darcy’s law as a porous medium charac-terized by porosity and permeability. It was shown that highly accurate CFD modelscould be engineered based on real scaffold geometries obtained from μCT scans topredict local mechanical effects even at single cell level [25]. The SF scaffolds usedfor all experiments presented in this thesis are polymeric scaffolds exhibiting a lowdensity. When immersed in culture medium and scanned with μCT the scaffoldcannot be distinguished from the culture medium because their densities are toosimilar. It is possible to scan SF scaffolds dry in air, to improve the contrast, never-theless it is known that SF scaffolds show swelling when immersed into a liquid [26].Therefore, the dry structure does not fully represent the wet structure. There isone study that determined SS in the full volume of a perfused SF scaffold. Thescaffold structure was acquired in dry state using μCT and the swelling was simu-lated by reducing local curvature by image post-processing [20]. Implementing real

136

scaffold geometries into computational models always leads to a massive increasein computational time. CFD studies have been performed on real scaffold geome-tries, but most of them were only able to show results in a small volume of interestdue to computational limitations [16, 27]. These limitations could be overcome bysimulating regular scaffold geometries fabricated by highly reproducible techniqueslike 3-D printing. Regular scaffolds however, do not resemble the stochastic poredistribution of bone and might not be an ideal solution for bone regeneration ap-plications. Another limitation of the CFD simulation is the lack of tissue growth.Tissue growth within the scaffold structure leads to decreased pore size and poros-ity, subsequently increasing flow velocities and SS acting on the cells. The fluidflow applied should therefore be adapted over the course of an experiment to keepthe mechanical stimulation of cells constant. This is not possible so far, becauseof various limitations. Non-mineralized extracellular matrix (ECM) growth and thelocation cells within the scaffold cannot be visualized with μCT due to their lowdensity. Other techniques like confocal microscopy could be implemented in thefuture to monitor non-mineralized ECM formation but this would require furtheradaptation of bioreactor design if mineralized and non-mineralized ECM of the samesample would be monitored simultaneously. Flow velocities within a heterogeneousscaffold structure are inherently highly variable and have been shown to vary bya factor of up to 2000 [11, 28]. With increased mineralized tissue formation cellswill be buried within the mineralized matrix like osteocytes in-vivo [29]. All theselimitations make it impossible to adapt the fluid flow velocity to match optimalmechanical loading conditions for every cell on the scaffold.

The in-vitro studies presented in chapter 3 and chapter 4 have been performedusing hMSCs. In-vivo, osteocytes are known to be the mechanosensitive cells andto stimulate bone formation and resorption, but it is known that stem cells aremechanosensitive as well. Various in-vitro studies have shown mechanically inducedosteogenic differentiation of hMSCs [10, 11]. The in-vitro studies performed in thisthesis were not compared to cell cultures using differentiated cells and no conclusioncan be drawn if the experiments would have been improved in terms of mineralizedtissue formation for example when using osteoblasts. Nevertheless, compared todifferentiated cells, hMSCs exhibit several advantages. They can be easily isolatedfrom various types of tissues like umbilical cord, bone marrow, adipose tissue etc.with minor donor site morbidity whereas differentiated cells can only be harvestedfrom bone tissue itself, leading to the same drawbacks as observed for current boneregeneration techniques. Stem cells have the ability of self-renewal and exhibit ahigh proliferation potential, which makes them suitable for in-vitro expansion [8].

137

Chapter 5 Synthesis

Differentiated cells on the other hand have a very low proliferation potential [30].Taken all these advantages together, it can be concluded that hMSCs might be thebetter cell source for clinical BTE applications with autologous cells [31, 32].

The mechanical loading regimes investigated in chapter 3.3 were chosen basedon SS calculated using a simple cylindrical pore model [33] and in agreement withvalues reported from literature to either induce proliferation or differentiation ofhMSCs [10, 34]. In chapter 4 it was shown that the mechanical loading regimedefined for the differentiation of hMSCs did not improve the amount of mineralizedtissue formation compared to static samples. This leads to the assumption thatthe loading regime chosen is not optimal to enhance mineralized tissue formationcompared to static cultures. Nevertheless, the aim of this thesis was to understandcausal relations between influencing parameters, such as mechanical environment,and experimental outcomes rather than to increase and optimize mineralized tissueformation in perfusion bioreactors. The increase of mineralized tissue formationcompared to static cultures was of little relevance. For future BTE experiments it isimportant to aim for functional constructs matching the geometrical and mechanicalfeatures of bone in-vivo and to include other cell types like endothelial cells forvascular growth or osteoclasts and osteoblasts for bone remodelling. Nevertheless,it is not clear up until now which parameters are necessary to produce a tissueconstruct in order to be functional after implantation [22].

SF scaffolds are widely used in-vivo and in-vitro [35, 36]. SF is known for its ex-cellent biocompatibility, controllable degradation and favourable mechanical prop-erties [1, 37, 38]. Despite these advantages, one major concern is the reproducibleproduction of SF scaffolds geometries. In chapter 4, this was an important limita-tion. The outer dimensions of SF scaffolds were produced using a precision saw anda cork borer. The channels were introduced with biopsy punches, but they could notbe produced equally leading to not completely round channels and distorted outerscaffold geometries. In wet state, SF scaffolds were soft and flexible. They were easyto cut, but also easily compressed during cutting possibly leading to the differencesin geometries observed. In dry state, SF scaffolds were hard but very brittle andit was not possible to cut them without breaking them. The limitation of exactreproduction of scaffold and channel geometries needs to be considered especiallyfor perfusion cultures. If the outer dimensions of the scaffold are too small it cannotbe tightly fit into the bioreactor and the perfusion flow would flow around and notthrough the scaffold leading to changes of the mechanical loading experienced bythe cells.

In the past few years perfusion bioreactors have been applied in numerous BTE

138

engineering studies showing positive effects on osteogenic tissue formation. Influenc-ing parameters like mechanical loading or culture medium composition however, arestill poorly understood and mostly determined by trial and error. The implemen-tation of computational simulations in combination with μCT helps to understandthe causal relations between influencing parameters and experimental outcomes andcan be used to optimize bioreactor design and culture conditions. In the future,the perfusion bioreactor system developed could serve as a framework to investigatedifferent influencing parameters of BTE cultures.

Conclusions

In conclusion, the development of the perfusion bioreactor system was completedsuccessfully by the investigation of the effect of curvature on 3-D mineralized tis-sue formation under loaded and static conditions. The original perfusion bioreactordesign was adapted leading to a homogeneous mechanical environment within thebioreactor and the scaffold. Cell culture conditions, namely culture medium sup-plementation with FBS and mechanical loading regime, were optimized and definedfor future experiments. The perfusion bioreactor system developed and the knowl-edge gained from optimizing culture conditions were finally combined and applied toinvestigate the influence of curvature on 3-D mineralized tissue formation. We be-lieve that the system developed during this thesis offers a framework to understandcausal relations between different influencing parameters and biological processes inBTE applications. Ultimately, the understanding of these relations could be usedto guide the development and to increase the quality of novel strategies for bonereplacement.

References

[1] C. G. Finkemeier. Bone-grafting and bone-graft substitutes. J Bone Joint SurgAm, 84-A(3):454–64, 2002.



139

Chapter 5 Synthesis






[9] M. Liebschner and M. Wettergreen. Topics in tissue engineering, chapter Op-timization of Bone Scaffold Engineering for Load Bearing Applications. 2003.







140





[20] E. Zermatten et al. Micro-computed tomography based computational fluiddynamics for the determination of shear stresses in scaffolds within a perfusionbioreactor. Ann Biomed Eng, 42(5):1085–1094, 2014.

[21] C. E. A. Jochems, J. B. F. van der Valk, F. R. Stafleu, V. Baumans. The useof fetal bovine serum: Ethical or scientific problem? Atla-Altern Lab Anim,30(2):219–227, 2002.

[22] F. Guilak, D. L. Butler, S. A. Goldstein, F. P. T. Baaijens. Biomechanics andmechanobiology in functional tissue engineering. J Biomech, 47(9):1933–1940,2014.

[23] K. Bieback et al. Human alternatives to fetal bovine serum for the expansionof mesenchymal stromal cells from bone marrow. Stem Cells, 27(9):2331–41,2009.

[24] H. Jinnai et al. Surface curvatures of trabecular bone microarchitecture. Bone,30(1):191–194, 2002.

[25] A. Boccaccio et al. Finite element method (FEM), mechanobiology andbiomimetic scaffolds in bone tissue engineering. Int J Biol Sci, 7(1):112–132,2011.

[26] U. J. Kim, J. Park, H. J. Kim, M. Wada, D. L. Kaplan. Three-dimensional aqueous-derived biomaterial scaffolds from silk fibroin. Bioma-terials, 26(15):2775–2785, 2005.

141

Chapter 5 Synthesis

[27] D. Lacroix, A. Chateau, M. P. Ginebra, J. A. Planell. Micro-finite elementmodels of bone tissue-engineering scaffolds. Biomaterials, 27(30):5326–34, 2006.

[28] J. Vance, S. Galley, D. F. Liu, S. W. Donahue. Mechanical stimulationof MC3T3 osteoblastic cells in a bone tissue-engineering bioreactor enhancesprostaglandin E2 release. Tissue Eng, 11(11-12):1832–9, 2005.



[31] D. C. Ding, W. C. Shyu, S. Z. Lin. Mesenchymal stem cells. Cell Transplant,20(1):5–14, 2011.

[32] A. I. Caplan. Mesenchymal stem-cells. J Orthop Res, 9(5):641–650, 1991.


[34] D. Li, K. Dai, T. Tang. Effects of dextran on proliferation and osteogenicdifferentiation of human bone marrow-derived mesenchymal stromal cells. Cy-totherapy, 10(6):587–596, 2008.

[35] L. Meinel et al. Engineering bone-like tissue in vitro using human bone marrowstem cells and silk scaffolds. J Biomed Mater Res A, 71(1):25–34, 2004.




142

perfusion bioreactors for bone tissue engineering - a combined

Documents