Professional Documents
Culture Documents
DEPARTMENT OF BIOENGINEERING
JOSHUA D. SALVI
Spring 2009
A thesis
submitted in partial fulfillment
of the requirements
for a baccalaureate degree
in Bioengineering
with honors in Bioengineering
Henry J. Donahue
Baker Professor of Cellular and Molecular Physiology
Thesis Co-Supervisor
Peter J. Butler
Associate Professor of Bioengineering
Thesis Co-Supervisor
William O. Hancock
Associate Professor of Bioengineering
Honors Adviser
Date of Signature
_____________________________________ ______________
Henry J. Donahue
Baker Professor of Cellular and Molecular Physiology
Thesis Co-Supervisor
_____________________________________ ______________
Peter J. Butler
Associate Professor of Bioengineering
Thesis Co-Supervisor
_____________________________________ ______________
William O. Hancock
Associate Professor of Bioengineering
Honors Adviser
9-6939-2728
ABSTRACT
Continued use of autografts and allografts for bone-tissue therapy has serious
implications, including the decreased strength of such grafts in vivo over time [2]. A
novel solution to this problem is the use of tissue-engineered implants from the patient’s
bone marrow. Numerous techniques have already been used to build either two- or three-
fluid flow-induced shear stresses will aid in the chemotransport among osteoblastic
been analyzed with their flat counterparts under both static and flow conditions.
Furthermore, these protocols have been simulated through finite element analysis in
COMSOL by solving for various stresses encountered by cells under oscillating fluid
progenitor cells, including human mesenchymal stem cells (hMSCs) and human bone
marrow stromal cells (hBMSCs), have been proliferated while maintaining their
differentiation potential into the osteoblastic lineage in vitro. These studies culminated in
the design of a three-dimensional bioreactor in which finite element analysis was used to
optimize the stress distribution and perfusion throughout the volume of a scaffold. The
purpose of these experiments was to explore the field of tissue engineering as it applies to
the proposal of a new in silico method and an analysis of key parameters for successful in
Bone Grafts:
Ten years ago, there were 650,000 reported cases of bone allograft transplantation
[9]. Bone transplantation is key in rebuilding diseased tissues. These include the hips,
shoulders, knees, and spine. Grafts are also key in the repair of bone loss from fractures
or cancers [10]. Allografts are tissue transplantations between individuals of the same
species. In this case, bone is grafted from one human being to another. Autografts, on the
other hand, are grafts of tissue from one location on an individual to another location of
that same individual. Autografts are far less common than allografts, due mainly to the
autograft transplantation, and allografts are thus necessary. Xenografts, methods rarely
used in orthopedic surgery, involve tissue grafts from one species to another. Bone
allografts are the most common due to their osteoconductive properties and the inclusion
of osteogenic factors that induce bone tissue growth [9]. These factors include bone tissue
Allografts are harvested either from a living donor or cadaveric sources within 24
hours of death [9]. Key components of this harvesting include the maintenance of cell
viability and prevention of allograft infection. This is not to say that current methods are
perfect, but they have been successful nonetheless. Risk of HIV infection is very low, at a
rate of 1 in 150,471. If lymph nodes are tested prior to transplantation, this rate has been
reduced to as low as 1 in 1.67 million [9]. There have also been cases of HCV
transmission, though these were rare. Though current sterilization methods are not
6
perfect, grafts rarely lead to disease transmission. Nonetheless, a viable alternative
Other factors affect allograft performance. With continued use in vivo, allografts
lead to mechanical failure of bone. Using human models, it has recently been
(such as elasticity), and increased numbers of micro-fissures resulted with time. These
data were correlated to the 60% failure rate of allografts within 10 years after
transplantation [2].
younger in order to decrease failure rates. Furthermore, additional tissue processing for
sterility (e.g. freeze-thaw cycles) increase failure rates [9]. These data can be correlated to
Thus, current bone graft methods are limited by the lack of donors, possible
disease transmission, and known mechanical failure with time. A tissue engineering
method incorporating autograft principles but assuring the osteogenic potential would act
unique example can be found in nature. Amphibian limb models are classic in that they
demonstrate that a limb can be naturally regenerated after injury. The unique factor in
7
ectopic blastemas followed by ectopic limbs were witnessed within 60 days post-wound.
Furthermore, if a skin graft was added to the wound location, amphibians demonstrated a
the injury, the key factors in limb regeneration (as opposed to skin regeneration alone or
fibroblasts are prevalent. The three pathways include wound healing (as seen in humans),
progenitor cell lines was required in order to form the regeneration blastema. These
phenomena in limb regeneration remain mysterious by mechanism, but the key factor of
utilizing one’s own cells to regenerate tissue becomes the focal point of skeletal tissue
engineering [11].
in skeletal tissue engineering, specifically to combat current issues with bone grafts [9].
8
Among the many strategies in skeletal tissue engineering, there are three basic
Construct
where a graft or scaffold will stimulate the growth of skeletal tissue. The second method,
however, involves in vitro culture of progenitor cells, typically harvested from human
bone marrow. These aggregates are then injected into the patient without an
accompanying scaffold. Finally, the third and most popular approach uses a complete
three-dimensional model of bone or cartilage. However, the cells within the scaffold must
there are numerous biomaterial factors that can influence cell-surface interactions. Within
seconds to minutes, cellular fluid adhesion is affected by surface wettability, and protein
adhesion can be affected by local pH, ionic composition, and temperature. Cell
attachment is then influenced by van der Waal’s forces. Finally, hours later, cell adhesion
and spreading are affected by matrix proteins and cytoskeleton proteins [12]. Two key
composition (i.e. ESCA). However, the focus of the following studies is surface
9
roughness studies result in controversial conclusions. Lithographic methods and
numerous microscale substrates resulted in conflicting results, and further data are
Similar methods and progenitor cells from bone marrow have been involved in
structures similar to chondrocyte repair. Though the final product may vary greatly, it
was demonstrated that the same progenitor cell lines can be used for numerous
applications [15].
for skeletal tissue engineering. Prior to developing a bioreactor, multiple cell lines,
substrates, and growth factors must be analyzed for potential differentiation and
proliferation in vitro. Furthermore, the bioreactor design is not the end result. Bone tissue
constructs must then be analyzed through implantation studies prior to the development
10
When designing a bioreactor, one must consider how nutrients and growth factors
the bioreactor system must account for multiple scaffold types, cell types, and scaffold
dimensions. These reactors have included tissue culture flasks, agitated vessels, packed
beds, fluidized beds, and membrane bioreactors [9, 12, 13, 16-18]. A systems view of
Figure 3. A Systems View of the Ideal Bioreactor in Tissue Engineering. Note the number
11
Progenitor Cell Lines:
Stem cells are unique in that they can continue to proliferate for long periods of
times, are unspecialized, and can differentiate into multiple lineages. Human embryonic
stem cells (hESCs) are harvested from a blastocyst initiated by in vitro fertilization [19].
The focus of this study, however, is on adult stem cells. Specifically, human
mesenchymal stem cells (hMSCs) are harvested from umbilical cord blood, bone
marrow, or fat in somatic tissue. There are two types of adult stem cells. Hematopoietic
stem cells give rise to erythrocytes, leukocytes, and platelets in the blood. However,
human bone marrow stromal cells (hBMSCs) are non-hematopoietic, meaning that they
do not give rise to blood cells. Instead, stromal cells differentiate into osteoblasts,
which stromal cells harvested from one tissue type can differentiate into the lineage of
another phenotype. For example, bone marrow stromal cells can differentiate into bone,
cardiac muscle, and skeletal muscle lineages [20]. The differentiation pathways
12
Figure 4. Comparison of hematopoietic (Top) and stromal (Bottom) stem cell
differentiation [20].
tree, where continued growth leads to a fairly complex matrix of numerous branches.
This morphogenesis becomes key in the differentiation of progenitor cells, whereby the
In bone, the branching of tissue by the dynamic nature of osteoblastic growth and
osteoclastic decay leads to a lighter and more structurally sound system. In fact, such
properties also lead to improved diffusion across the mesenchyme [21, 22]. However,
13
Initially, pluripotent stem cells can differentiate to the myoblast, fibroblast,
commit as a progenitor cell that can then differentiate to all but the myoblastic and
differentiation potential to the osteoblastic lineage. These osteoblasts can then further
The major issue with tissue engineering, however, has been that continued
Tissue engineering techniques thus attempt to mimic the in vivo environment of bone in
order to maximize this differentiation [21, 24, 25]. This technique typically requires some
14
Finite Element Analysis:
engineers often turn to structural analyses by the finite element method (or finite element
analysis, FEA). FEA allows for the discretization of complex geometries, granting much
greater flexibility in the study of mechanics than finite volume methods (FVM). The key
difference between FEA and FVM is the analysis by nodes or volumes, respectively.
Typically, finite element analyses require the use of computer technology for discrete
approximations [26].
number of finite elements or nodes and then approximating between them. First, consider
any differential equation. For our purposes, let this be a second-order differential
, where L is some linear operator. Using dot products, we can then state the
then solve for this equation instead of solving for . This equation is known as
the Galerkin form, and it now requires that L only be a stationary point as opposed to a
linear operator. This method for solving equations by nodal approximations becomes key
FEMLAB) has been used. COMSOL is unique in that it can not only analyze complex
geometries by finite element analysis, but it can also couple multiple physics modules
15
into a single geometry. The program was initially developed by Germund Dahlquist at the
Royal Institute of Technology in Stockholm, Sweden [28]. This created the ability to
undergoes changes of mechanical properties and changes of relaxed lengths. These are
the use of multiphysics to study them simultaneously. By using microspin velocity as the
rate by which bone remodeling occurs, and noting both the current state and relaxed state
of each element, constitutive equations could then be coupled. Finally, these models were
able to find the anisotropic elasticity with respect to time. Though not described in detail
here, this simulation exemplifies an excellent application of the finite element method to
cells in vitro is mimicking the in vivo environment. Typically, cells are often cultured on
surprisingly, cells cultured on these substrates lose their differentiation potential with
continued expansion.
dimensional substrates that mimic the in vivo environment. The first method, the one
16
used in this particular
p stuudy, is an alteeration of phhysical surfaace topograpphy while hoolding
suurface chem
mistry consttant (Figure 6.a). The polymer iss arguably a more effiicient
m
morphology, cytoskeletaal organizattion, and protein
p adsorption at the cell-suurface
B
Biofunctional
lization allow
ws a firm ceell anchoragee through peeptide motifs
fs that can biind to
inntegrins on the
t cells. Paassive coatinng involves natural
n proteein adsorptioon to the suurface
(F
Figure 6.d). Additionallyy, these pepttides can bee covalently linked to the surface (F
Figure
6.e). Finally, peptide linkkers can be used to binnd these prooteins to the material suurface
(F
Figure 6.f). Typically, combination
c s of these biomimetic
b p
properties w
would exist in the
inndependently
y [31, 32].
17
Though demineralization improves allografts and autografts after implantation,
previously, these scaffolds provide a matrix upon which progenitor cells can grow into
mature bone. This construct is then re-implanted into the patient from whom bone
multitude of other materials and approaches have been used in the development of
scaffolds. However, it has been demonstrated recently that mineralization by calcium and
further growth of osteoblasts [33]. These properties thus become key in the development
of scaffolds.
Bioreactors:
The bioreactor provides a sterile environment for in vitro cell culture. The goal of
the design is to mimic in vivo conditions, thus improving proliferation and differentiation
of mesenchymal stem cells. Skeletal tissue engineering bioreactors include those with
fluid flow or mechanical stretching [34-37]. These provide dynamic mechanisms for the
In a myriad of past studies, static cultures were involved. Though static cultures
scaffolds. This phenomenon is expected, since cells in vivo are grown under mechanical
18
loads by interstitial fluid flow [34]. As a result, implementation of biophysical signals
Spinner flask bioreactors, on the other hand, use a magnetic stir bar to thoroughly
mix the media. Cells cultured in these bioreactors demonstrated increased proliferation
and uniformity of distribution. However, the spinner flask bioreactor was not the optimal
device for in vitro osteogenesis due to decreased cell adhesion under turbulent flow
patterns [34].
implements two concentric cylinders, with the outer cylinder rotating. Scaffolds then
float in the annular space after centrifugal forces and gravitational forces are balanced.
engineering. However, the rotating wall vessel bioreactor does not significantly induce
osteogenesis. One factor discovered by NASA, though, was that turbulence in bioreactors
properties significantly reduce cell activity. Thus, this bioreactor functions through
Flow perfusion bioreactors are models very similar to the one developed in this
study. These devices perfuse media through scaffolds with a pump mechanism and can be
perfusion bioreactors improve mass transport throughout the interior of the scaffold,
unlike the spinner flask and rotating wall vessel models. Furthermore, fluid shear stress
has been demonstrated to induce osteogenesis in vitro. Studies comparing static and
19
perfused scaffolds in a flow perfusion bioreactor found that calcium deposition
significantly improved in these devices. Since interstitial fluid flow is a key component of
in vivo bone growth, this bioreactor acts as a significant improvement over the
mass transports of nutrients throughout the interstitial space of the scaffold. For example,
bioreactors have induced mechanical loading by oscillating fluid flow, substrate bending,
20
Literature Review:
Two-Dimensiional Substraata:
T
This method provides a uniform texxture upon which
w cell morphology
m a adhesionn can
and
be readily con
ntrolled. Typpically, a sillicon wafer or
o similar siilicon substraate is the staarting
m
material. Pho
otoresist is then addedd to the waafer by spinn casting. Ultraviolet
U (UV)
faabricated, th
he substrate can then bee casted witth a numberr of materiaals. For exam
mple,
caalcium or po
olymers suchh as polylacctic acid (PL
LA) or polym
methylmethaacrylate (PM
MMA)
arre used as co
ommon materials [41-444]. This metthod is summ
marized grapphically in Figure
F
7.
Figure 7. Fabrication
F of Two-Dim
mensional Suubstrata by Photolithogr
P raphy [42, 433].
The major
m issue with
w photolitthographic methods
m is thhat these subbstrata encom
mpass
21
m
milieu of bo
one [1, 45]. Thus, randdomly distriibuted topoggraphies maay provide more
Figure 8. A comparison
c of multiple Figure 9. hFOB B area and adhesion
demixxing ratios. [1] after culture onn various PLLA//PS
demixed substrata
s [1].
crreated throu
ugh a polym
mer demixinng method. In polymerr demixing, two immisscible
evvaporate. Th
he polymers will then seegregate intoo two phasess, and varyinng concentraations
suubstrates [46
6].
22
Upon using the polymer demixing method, it was found that various
weight/weight ratios would result in very different topographical features. When poly(L-
result in a topography more similar to nanopits rather than nanoislands (Figure 8). After
demixing, poly(L-lactic acid) would segregate to the upper surface of the polymer film,
mimicking similar culture dishes. After analysis by XPS and contact angle (wettability),
it was found that the PLLA/PS 70/30 w/w resulted in the most in vivo like conditions [1].
When human fetal osteoblastic cells (hFOB) were cultured on each of these
surfaces, both cell morphology and adhesion were measured. Additionally, these
substrata were compared with both flat poly(L-lactic acid) and polystyrene surfaces. In
doing so, it was thus determined that cell area was significantly greater (p < 0.01) on both
the poly(L-lactic acid)/polystyrene 70/30 and 90/10 w/w substrata when compared with
the flat controls. Furthermore, cell adhesion was also significantly greater (p < 0.01) on
these same substrata when compared with the flat controls. These demixed films
correspond to nanoislands as opposed to the nanopits illustrated in Figure 9.b. These data
demonstrate that osteoblastic cells will preferentially grow and adhere to various surfaces
based upon surface topography alone. This conclusion corresponds well with tissue
that nanoscale surface topography positively influences cell morphology and adhesion
23
when other biomaterials were used [1, 45, 47-52]. The key issue, however, was that cell
activity and both proliferation and differentiation potentials were not measured in these
maintenance of differentiation potential in vitro are the true keys to skeletal tissue
engineering. Without such data acquisition, it cannot yet be concluded that these substrata
are in fact beneficial to osteoblastic differentiation of human bone marrow stromal cells
However, other studies have demonstrated that inducing biophysical signals via
oscillating fluid flow may increase proliferation and possibly differentiation potential.
The data in support of this theory simply used flat substrates as opposed to the
upregulation of the MAP kinases ERK 1/2 result from oscillating fluid flow-induced
shear stresses [8, 39, 53-56]. Again, polymer demixed films were not involved in these
experiments. It should also be noted that a similar study analyzed the elastic moduli of
displayed significantly greater elastic moduli when compared with those on flat
polystyrene or plasma-cleaned glass surfaces [57]. Intuition would then beg the
hypothesis that some combination of induced biophysical signals and surface topography
would result in increased apparent shear stresses due to the increased elastic moduli of
cells cultured on these substrata. Thus, such a combination may in fact provide a more
versatile and adept method for maintenance of stem cell differentiation potential with
24
Three-Dimensional Scaffolds:
has been widely disputed that three-dimensional scaffolds provide a more adept
microenvironment due to increased surface area and a better matrix for bone growth in
skeletal tissue engineering. Numerous studies have analyzed such scaffolds, but none of
measure along with induced biophysical signals to promote progenitor cell activity [4].
PEG3400-RGDS (C), and methacrylated heparin (D) used for hydrogel fabrication [4].
necessary in the commitment of progenitor cells to the osteoblastic lineages (Figure 5).
Recent studies in multiple laboratories have used PEG hydrogels to create a scaffold via
25
copolymerization (Figure 10). Note that the structures of these polymers have a high
molecular weight and a high density of ester functional groups. These become key in the
interaction with BMP2. The PEG and heparin copolymers noncovalently bonded to the
BMP2, resulting in a more adept method for controlling a sustained level of osteogenic
growth factors within the scaffold [4]. As a result, these data were correlated to increased
differentiation potential of progenitor cells into the osteoblastic lineage in vitro on these
three-dimensional hydrogels.
However, hydrogels can be expensive to fabricate and maintain, and other simpler
scaffolds have also been thoroughly explored. Another example was the development of
extracellular matrix of bone. Scanning electron microscopy images and pore sizes of
frozen at -30°C (C) and -80°C (D), along with porosities (E).
26
Note that most pores were on the order of 50 to 200 microns, near the range of
marrow stromal cells (approximately 20-40 microns). After culturing osteoblastic cells on
osteoconduction both in vitro and in vivo [5]. However, these scaffolds must be stored at
concept of utilizing physiologic biomaterials will be applied in this study, but a scaffold
scaffolds.
been popular recently, but these scaffolds tend to have very little three-dimensional
depth; in fact, it is argued by this investigator that these fiber networks resemble more of
progenitor cell proliferation when compared with two-dimensional tissue culture plates.
networks when compared with the two-dimensional plates. Again, note that these plates
were basic tissue culture substrates, and the networks were not compared with other
scaffolds nor with multiple surface topographies. Finally, osteocalcin was significantly
the extracellular matrix in skeletal tissue engineering. Though the nanofiber networks
provide a significant improvement over tissue culture plates, it must be noted again that
27
these networks are not truly three-dimensional scaffolds when compared with the other
aforementioned scaffolds.
Some studies delve into comparisons between scaffolds, but these comparisons
Poly(lactic acid) has been approved by the Food and Drug Administration for therapeutic
use due to its biodegradability. Typical quantification methods in these studies analyzed
alkaline phosphatase, osteopontin, and osteocalcin through assays [65]. However, all of
these were in static culture, and there was only comparison between commercially-
available three-dimensional scaffolds. None of the studies compared these scaffolds with
Many tissue engineers are now beginning to grasp the concept that cell seeding is
enhanced through the application of a bioreactor. One example is the use of poly(DL-
Cell seeding was significantly enhanced by the utilization of a spinner flask bioreactor
[3]. As mentioned previously, however, these bioreactors do not provide optimal fluid
polymer used in this method was also approved by the FDA, so a similar biomaterial
would thus be preferred in the development of the scaffold in this particular study.
28
Putting biomaterial properties aside, a key component in the fabrication of
optimal three-dimensional scaffolds is the mesh size preferred by osteoblastic cells and
their progenitors. Using a titanium fiber mesh, it was determined that mesh size and
porosity does, in fact, affect cell responses to growth on the scaffolds. Under both static
and flow conditions, alkaline phosphatase activity was significantly altered between 20
micron and 40 micron average pore sizes. Furthermore, mineralization and osteopontin
secretion were also significantly affected by mesh sizes in the corresponding scaffolds.
Figure 12.A depicts the titanium mesh after four days of culture. By day 8 in culture
(Figure 12.B), mesenchymal stem cell concentration began to increase. Finally, day 16
(Figure 12.C) of culture demonstrates a confluence of cells and mineralization within the
construct [66].
methods were used analyzing various biomaterials and porosities. However, studies have
29
yet to compare many types of scaffolds with their two-dimensional counterparts while
Bioreactor Designs:
One of the most common methods for seeding cells in three-dimensional scaffolds
is by the use of flow perfusion bioreactors. Each of these reactors are key for initial
scaffold seeding, but a bioreactor with oscillating fluid flow has yet to be devised.
Detailed below are a few of these bioreactor designs and their respective qualities.
unidirectional flow from a pump that will perfuse media with cells through the path of
Figure 13. Schematic of a standard flow perfusion bioreactor. Note that no flow occurs
around the outer perimeter.
The bioreactor detailed above simply involved the basic design of the reactor
system and peristaltic flow loop [3]. However, the seeding efficiency was not analyzed,
begging the question as to whether or not the aforementioned system was truly
advantageous over other similar bioreactor systems. Does the flow and seeding in a
perfusion bioreactor provide uniform seeding density? Does the modification of scaffold
30
porosity affect the perfusion or shear stresses throughout the scaffold while perfusing
media through the scaffold? These questions have not yet been answered by Bancroft in
Years prior to the design of the bioreactor by Bancroft, many tissue engineers
developed small flow chambers for two-dimensional substrata. In essence, these were
also similar in that they perfused media through the system. Furthermore, the study also
analyzed continuous unidirectional flow over bone marrow stromal cells in culture. Note
that this flow is not oscillatory as occurs in vivo, but the use of continuous flow in vitro
still provides some mechanism for improvement over static culture. Many flow loops
oscillatory fluid flow was rarely used in these studies [3, 67].
Using highly porous collagen microspheres for long term in vitro culture of bone
maintained in the bioreactor system under unidirectional peristaltic flow. In this particular
case, “long term” flow referred to culture of murine bone marrow cells over a four month
traditional static culture in flasks. The authors of the study pointed out, however, that
preferred. Future bioreactors should incorporate such maintenance, and other possibilities
also include an improvement of culture media or the growth factors involved in the
experiments [67].
31
chamber similar to that constructed in the previous study. The goal was to analyze the
effects of flow rates and then the capability for differentiation with time in vitro [7].
The scaffolds analyzed in these studies were PET matrices, which were in fact
dimensional alternatives. After setting up the flow loop, it was determined that there
exists an inverse relationship between flow rate and seeding density. This is to say that
lower flow rates result in much greater seeding densities. However, cells did not begin to
proliferate significantly until after the first four weeks had passed, as depicted in Figure
14 below [7].
Figure 14. Cell density seeded on three PET scaffolds with time in a peristaltic flow
chamber [7].
The bioreactor studies with PET meshes demonstrated that faster seeding kinetics
32
into the osteoblastic lineage was significantly enhanced in these two-dimensional
biophysical signals, and primary donor cells were not used in these experiments. Such a
lack of analysis leaves a large gap open for study when compared with numerous studies
Nonetheless, this perfusion bioreactor system was quite successful in its attempt
to culture cells to confluence and induce differentiation into the osteoblastic lineage. An
example of the PET mesh after culture in the perfusion bioreactor is detailed in Figure 15
below. After 40 days of culture, a dense cellular mass was witnessed under scanning
electron microscopy. It can thus be determined that bioreactor systems provide numerous
Figure 15. SEM image of PET scaffold with hMSCs cultured for 40 days [7].
As can be seen, numerous flow perfusion bioreactor systems have been
oscillatory flow occurring in vivo. Additionally, many bioreactors tend to implement two-
33
biophysical and biochemical signals in vitro, a novel bioreactor design would better
mimic the in vivo environment through oscillatory flow and three-dimensional constructs.
physical properties in more complex geometries. In tissue engineering, these analyses are
flask bioreactors and perfusion flow bioreactors. However, fluid flow over cells cultured
Regarding the simulation of flow perfusion bioreactors, the most common method
into a number of nodes. The Lattice-Boltzmann method is very similar to the method
described previously. The important factor to consider, however, is that fluid flow
second-order set of equations and assume that all fluids are Newtonian fluids. As a result,
the calculations with these methods are very rough estimates of the reality at hand [79].
fields through bioreactors, and shear stresses were coupled with velocities at the solid-
fluid interface. In coupling empirical data regarding scaffold properties and then
analyzing flow fields, the flow field properties existing within the flow perfusion
below. The upper image depicts the transverse velocity field through a three-dimensional
scaffold in mm/s. The simulation image on the bottom of Figure 16 displays the same
34
three-dimensional scaffold, but with a top view [79]. Note, again, that the velocity
Figure 16. Velocity fields in a flow perfusion bioreactor as estimated by the LB method.
The upper image depicts transverse flow through a scaffold from a side view, and the
bottom image depicts the same flow from a top view [79].
Finite element analyses become key in the design of an optimal bioreactor system.
Many tissue engineers have failed to perform these simulations prior to bioreactor design,
much unlike the approach taken in this study. Furthermore, simulations have been limited
35
Literature Review Summary:
The table below is meant to compare many of the current techniques used in bone-tissue engineering. Notice that a single study
does not use both 2D and 3D substrates while coupling these with biophysical signals. Additionally, the use of computer simulations
(finite element analysis) is unique to this study. The proposal listed on this table will be described in the next section.
Desired Biophysical
Author Biomaterial 2D/3D Cell Line Quantification Novel Aspects
Lineage Signals
Murine
CFU-GM
Wang, ‘99 Bovine Collagen Bone Porous
2D Expansion No Assay,
[67] Scaffolds Marrow Microspheres
Hemacytometry
(C57BL/6J)
Bancroft,
poly(DL-lactic-co- Rat Marrow Hemacytometry,
‘03 3D Osteogenic No 3D Bioreactor
glycolic acid) Scaffolds Stromal PCR
[3]
ALP,
Holtorf, ‘05 Mesh Variability,
Titanium Fiber Mesh 3D hBMSC Osteogenic Yes Osteopontin,
[66] Constant Porosity
Ca2+ Deposition
poly(L-lactic Adhesion Adhesion, ALP,
Lim, ‘05 Nanoscale
acid)/polystyrene 2D hFOB and No SIMS,
[1], [45] Substrata
Demixed Substrata Expansion Hemacytometry
Porter, ‘05 Lattice- Computational
N/A 3D N/A N/A N/A
[79] Boltzman Modeling
36
ALP,
Mineralization,
Trojani, ‘05 RT-PCR Injectable Hydrogel,
(Si-HPMC)-based Hydrogel 3D Osteosarcoma Osteogenic No
[6] (TGFβ, pH Maintenance
Interleukin,
etc.)
Zhao, ‘05 poly(ethylene terepthalate)
3D hMSC Osteogenic No Cell Density Perfusion Bioreactor
[7] Mesh
Scaffold Dry
Human Comparison of
Hee, ‘06 poly(lactic acid) Foam, Beta Weight, ALP,
3D Dermal Osteogenic No Commercially
[65] TCP Scaffolds Histology,
Fibroblasts Available Scaffolds
Osteopontin
Hosseinkhani,
Peptide-Amphiphile ALP, “Three-Dimensional”
‘06 “3D” hMSC Osteogenic No
Nanofibers Osteocalcin Nanofibers
[61]
Riddle, ‘06 Intracellular Application of
Glass Slides 2D hMSC Expansion Yes
[55] Calcium Mechanical Loading
Bidirectional
Benoit, ‘07 poly(ethylene glycol)
3D hMSC Osteogenic No BMP2, ALP Interaction between
[4] Hydrogels
Cells and Hydrogel
Cell Stiffness
Hansen, ‘07 polystyrene/polybromostyrene Cellular
2D MC3T3-E1 Expansion No correlated with
[57] Demixed Substrata Modulus
Nanoscale Substrata
poly(L-lactic Adhesion
Lim, ‘07 FAK, pY397, Adhesion on
acid)/polystyrene 2D hFOB and No
[50] Expansion Integrin Nanoscale Substrata
Demixed Substrata
ATP, Western
Blot,
Riddle, ‘07 Fluid Flow-Induced
Glass Slides 2D hBMSC Expansion Yes Intracellular
[56] Calcium, Proliferation
Calcineurin
37
Histology,
Dawson, ‘08 Osteogenic, Microchannels
Type I Collagen Scaffolds 3D hBMSC Chondrogenic No micro-CT,
[5] Promote Osteogenesis
ALP
1. Comparison of
1. PLLA/PS Demixed 2D and 3D
Substrata AFM, SEM, Protocols
2. PS/PBrS Demixed ALP, Flow 2. Design of a 3D
Substrata Yes Cytometry, Perfusion
2D hMSC (Empirical
PROPOSED 3. Calcium Phosphate Osteogenic Intracellular Bioreactor with
3D hBMSC and
Scaffolds Simulated) Calcium, Biophysical
4. PLLA Salt Leached Finite Element Signals
Scaffolds Analysis 3. Simulations of
5. Simulated Bioreactor Oscillating
Fluid Flow
38
Proposal:
Two-Dimensional Substrata:
This study will involve the fabrication and analysis of two-dimensional nanoscale
substrates with variations in surface topography. Specifically, the goal will be to analyze
To accomplish this polymer demixing will be used with the following immiscible
polymers:
Finally, the surface chemistry will be maintained among these substrates. These
will then be compared with flat controls of the same chemistry. In this case, either
Three-Dimensional Scaffolds:
Multiple scaffold variations will then be analyzed and compared with the results
progenitor cell activity will be compared both between the scaffolds and the
39
3. Salt-Leached Poly(L-lactic acid) polymer scaffolds (150-710 micron NaCl
crystals)
It should be noted that the wide range of salt crystals will be further divided into
relative sizes to obtain hypothetical variations in porosities. These scaffolds will vary
greatly in surface chemistry, surface topography, and relative porosities. For this reason,
bioreactor.
Biophysical Stimuli:
Bone in vivo undergoes biophysical stimuli by oscillating fluid flow through the
insterstitial space. Thus, an important factor will be the addition of these stimuli to the
Using oscillating fluid flow, shear stresses will be induced within the physiologic
range. Specifically, stresses will be induced at 5, 10, and 20 dyne/cm2. This flow will be
coupled with the two dimensional substrata only due to the lack of a current bioreactor
system for such flow. However, simulations of the bioreactor system being developed
will determine flow rates necessary to induce such shear stresses throughout the volume
of each scaffold.
Bioreactor Design:
A bioreactor will not be fabricated in this study, but a design will be created for
potential fabrication at a later date. Such a design involves certain criteria to be met. The
40
bioreactor system will be used for long-term culture of human bone marrow stromal cells
for differentiation into the osteoblastic lineage in vitro. All of the prior studies eventually
lead up to the design of such a bioreactor, so these criteria cannot be taken lightly. The
Standard scaffolds must fit within the Ensure the bioreactor has a variable
bioreactor volume. diameter between 2 and 10 mm.
aforementioned bioreactor. The above specifications must be met, and finite element
analyses will determine the optimal geometry, flow rates, and scaffold properties key in
Additionally, this study will simulate fluid flow of media over cells cultured on
the nanoscale substrata described previously. The goal will be to analyze the biophysical
41
determine where turbulence may result and what these flow patterns look like. In essence,
42
Thesis Statement and Hypotheses:
Thesis Statement:
Through a number of experimental protocols, skeletal tissue engineering methods
will be modified and made optimal (i.e., increased growth and differentiation) for
maintenance of stem cell proliferation and differentiation in vitro by mimicking
the in vivo extracellular milieu of bone and induced biophysical signals.
Hypothesis 1:
Hypothesis 2:
The same progenitor cell lines cultured on three‐dimensional calcium phosphate
scaffolds will display significantly maintained differentiation potential with
continued expansion in vitro compared with two‐dimensional substrata and flat
controls.
Hypothesis 3:
A bioreactor designed with finite element methods will satisfy all the criteria
needed for long‐term culture on three‐dimensional scaffolds with induced
biophysical signals.
43
MATERIALS AND METHODS
Nanoscale Substrate Fabrication:
Two methods for polymer demixing have been demonstrated to result in the same
(PBrS), with a molecular weight of 65x103 amu, and polystyrene (PS), with a molecular
weight of 289x103 amu, blend solutions were involved in one method. Another blend
order to assure nanoisland topography, PS/PBrS was blended in 60/40 w/w, and
PLLA/PS blended in 70/30 w/w. The blended, immiscible polymers were then dissolved
Once blended, the polymer solutions were then dispensed onto quartz slides or
glass cover slips depending upon the particular protocol. Quartz slides were used for flow
experiments, and cover slips were involved in numerous assays, including alkaline
phosphatase assays. These were then spin casted at 4000 rpm for 30 seconds. In doing so,
the volatile chloroform rapidly evaporated, allowing segregation of the two blended
polymers. The films were sealed and allowed to dry for 24 hours.
For the PS/PBrS polymer films, an annealing process was required to assure
uniform surface chemistry. The spin-cast films were heated to the glass transition
temperature (Tg) of PS but below the Tg of PBrS. Film concentrations of 0.5% and 1.0%
were annealed for 1 hour, and spin-cast films of 2.0% and 3.0% were annealed for 2
44
hours. The annealing process allowed the PS composition of the films to segregate to the
air-film interface.
After the fabrication process was complete, substrates were sealed and allowed to
cool for 24 hours if annealing was performed. Annealing was not required for PLLA/PS
spin-cast films. In preparation for cell culture, all substrates were exposed to UV
various diameters. The diameters were divided into three categories: 150-300 μm, 300-
500 μm, and 500-710 μm. The mixture was poured into a Petri dish such that the liquid
was approximately 3-5 mm deep. It was then sealed and allowed to dry for 48 hours.
The solid PLLA/NaCl scaffold was then leached three times with ddH2O in order
to dissolve and thus remove all NaCl crystals from the PLLA scaffold interface.
the scaffolds. The salt-leached PLLA scaffolds were then cut into cylindrical scaffolds
Becton-Dickson©. These scaffolds were then cut into cylinders that were 4 mm thick and
45
To prepare the scaffolds for cell culture, all scaffolds were immersed in 100%
EtOH for 24 hours. The ethanol was then aspirated and all scaffolds were exposed to UV
radiation for 60 minutes. This process was repeated three times to ensure complete
Cell Harvesting:
Human mesenchymal stem cells (hMSCs) were obtained from Cambrex©. These
Another cell line included primary human bone marrow stromal cells (hBMSCs),
another type of progenitor stem cell. To harvest these cells, human bone marrow was first
obtained from a patient at the Milton S. Hershey Medical Center through a protocol
defined by the Internal Review Board (IRB). The marrow was harvested from the rimings
of the femoral head of a 43 year old male patient undergoing hip surgery. Cells were then
washed and separated using ficoll gradient (1.077 g/ml). Cells located at the interphase of
this gradient were collected. They were then plated at a density of 2x105 cells/cm2 in
(DMEM), 10% fetal bovine serum (FBS), 1% Penn/Strep, and 1% L-glutamine. After
four days of incubation, non-adherent cells were removed and adherent cells maintained
46
Cell Culture:
hMSCs and hBMSCs were maintained in low glucose DMEM, 10% FBS, 1%
FBS, 1% P/S, and 1% L-glutamine until required for experimental protocols, with media
changes every 2-3 days. When cells reached a confluence of 70-90%, they were then
trypsinized and passed to prevent the natural differentiation that occurs when progenitor
cells reach confluence. The hMSCs were used up to a passage of 7-8, and the hBMSCs
were viable up to a passage of 4-5. When ready for data acquisition, cells were then
seeded onto the corresponding substrates or scaffolds with either growth or osteogenic
media used. Osteogenic differentiation media consisted of low glucose DMEM, 50 µg/ml
Cells were then seeded onto two dimensional nanoscale substrates at a density of
4x103 cells/cm2. For three dimensional scaffolds, cells were seeded at 5x106 cells/cm3.
This seeding would occur after sterilization of the appropriate substrates or scaffolds.
Scaffold seeding was performed via the vacuum filtration method. Substrate seeding was
Cells were then allowed to expand, proliferate, and differentiate for a period
defined by the particular protocol. Media changes took place every 2-3 days, and cells
47
Oscillating Fluid Flow:
After hMSCs were cultured on various nanoscale substrata, they were then
cleaned with PBS three times and loaded onto a vacuum flow chamber (Figure 17). Clear
flow media was perfused through the chamber, and all were attached to the pneumatic
correlated to shear stresses acting upon the substrata by the following equation:
, where τw is the shear stress at the wall, μ is the dynamic viscosity of the flow
media, Q is the amplitude of the sinusoidal flow rate, W is the width of the flow chamber,
and H is the height of the flow chamber. Since the dynamic viscosity, width of the
chamber, and height of the chamber were constant, the shear stress at the wall could then
48
be controlled by adjusting the amplitude of the sinusoidal flow rate. This is assuming
Poiseuille, laminar flow rates in a Newtonian fluid. In doing so, the pneumatic pump was
then used to induce shear stresses by oscillating fluid flow at 5, 10, and 20 dyne/cm2.
The fluid flow protocol called for the flow chamber to be coupled with a
fluorescent microscope calibrated for Fura-2 fluorescent markers. The first 180 seconds
were under static conditions, followed by 180 seconds of fluid flow. The entire period of
Fluorescent Markers:
For short-term oscillating fluid flow experiments, Fura Red fluorescent dyes were
loaded for 30 minutes prior to data collection. Fura 2 was not involved due to the fact that
absolute fluorescence was recorded for 360 seconds, the data was outputted to a PC. Cells
were manually selected, and each experiment included 15-40 cells within the data range.
Data processing included an initial average (μ) of absolute fluorescence during the
static period (0-180 s). During the flow period, the peak fluorescence was then
determined. If the peak fluorescence surpassed the average plus four times the standard
deviation of static fluorescence (μ+4σ), the cell was considered to have responded. This
same analysis was performed for each manually selected cell within each short-term
49
hBMSCs cultured for either 7 or 12 days in both growth medium and osteogenic
differentiation media were harvested and then tagged with SSEA-4, CD73 (SH3/4),
human embryonic stem cells has been shown to identify stem cells that are osteogenic in
vivo [80]; CD73 and CD105 react with bone progenitor cells, but not with osteoblasts or
These cells were then washed three times with PBS and suspended for
fluorescence-activated cell sorting (FACS) analysis. This analysis first uses flow
cytometry to sort cells by size and complexity. After sorting, the fluorescence of cells
was then measured and correlated with each of the antibodies. Each experiment analyzed
10,000 counts, which resulted in 7,000-9,500 cells per experiment on average. Controls
were used to determine the fluorescence of nonresponding cells, and a threshold was
determined for each experimental protocol in order to determine the percentage of cells
50
Alkaline Phosphatase Assay:
(ALP) assay is key in measuring the activity of this enzyme, which can then be correlated
An alkaline phosphatase assay kit was involved in this protocol. Two previously
prepared standards with 0.9% NaCl and blanks were first created to develop a standard
calibration curve for optical density and absorbance. Human mesenchymal stem cells
cultured for 12 days on two dimensional substrata and three dimensional scaffolds were
then lysed. The cell lysate was added to 65 mM phenolphthalein monophosphate in 7.8 M
To this mixture, 0.1 M phosphate buffer with a pH of 11.2 was added. The
mixture was diluted progressively in order to create a curve for each of the cell lysates.
The phosphate buffer acts as a color stabilizer when measuring the alkaline phosphatase
activity for the cells cultured on each of the substrates. These assays were then automated
in order to measure absorbance relative to blank samples and calibration curves. The
51
Substrate Characterization:
After fabrication of the two dimensional nanoscale substrata, the samples were
dried and prepared for atomic force microscopy (AFM). AFM was key in the analysis of
the nanoscale topography of these nanoislands. Specifically, the method allowed for
characterization of the scales and random distribution of the rough surfaces of these
The three dimensional scaffolds were not characterized utilizing AFM. Instead,
scanning electron microscopy (SEM) was used to characterize the three dimensional
characteristics and pore sizes among each of the scaffolds. SEM required that all
scaffolds were dried and coated with gold (Au) prior to analysis. Scanning electron
microscopy images were then outputted to ImageJ in order to determine the relative
porosities between the scaffold types. The software considered the relative white-to-black
balance along each of the upper surfaces of the scaffolds. This balance was then
correlated with the relative pore sizes along the scaffold. Porosity is normally defined by
the following equation: , where Vv is the volume of the void space, VT is the total
volume of the bulk material, and the porosity is the ratio between the two. Since SEM
space, AT is the total area of the surface, and estimated porosity is the ratio between these
pixels after the image undergoes a threshold algorithm, and AT is redefined by , which
is the total number of pixels in the image. This definition of porosity acts as a very rough
estimate and should only be considered as a relative porosity measure used to compare
52
scaffolds. The absolute value is, in essence, inaccurate due to the assumption that these
pores remain continuous throughout the bulk of the material, which is not true in reality.
Nevertheless, these relative values play key roles in the comparison between scaffolds.
surfaces of biomaterials plays a key role in cell adhesion and morphology. Decreased
of cells when analyzing their morphologies or histologies. Again, these data were
previously determined, and the measure of wettability simply ensured uniform surface
comparison were different levels of cell confluence, differing elastic moduli of cells
cultured on numerous substrata, and an analysis of the flow patterns over a cell cultured
on a hydrophobic substrate.
When comparing the cells cultured on multiple substrata, data from a contact-
mode atomic force microscopy (AFM) study by Joshua Hansen were used as inputs for
the moduli of cells on these surfaces. Cells cultured on plasma-cleaned glass had an
average modulus of 7000 Pa, those on flat polystyrene had a modulus of approximately
4000 Pa, cells on 11 nm nanoislands (PS/PBrS 40/60 w/w) displayed an elastic modulus
53
12000 Pa. Incompressible Navier-Stokes fluid flow was coupled with a stress-strain
analysis to determine the apparent shear stresses induced on the surfaces of these cells.
When analyzing oscillatory fluid flow, one must ensure that the flow remains
laminar if assumptions and equations are to hold. The maximum Reynolds number was
include:
•
•
•
These equations for fluid flow analyses can then be coupled with a stress-strain
In the analysis of shear stresses, the normalized von Mises stress was used:
1
6 6 6
√2
All of these data were then analyzed to determine flow patterns, values of shear
stresses at multiple locations, and the absolute flow rates in various geometries.
Note: COMSOL model reports detailing the inputs and outputs from the
54
Bioreactor Design:
In the design of the bioreactor, COMSOL Multiphysics was also used to create
various geometries and alter the flow rates. Porosity values previously determined were
used as inputs, with a permeability of zero, when simulating the insertion of a scaffold in
symmetry. In turn, this allowed the three-dimensional system to be analyzed through two-
dimensional simulations.
were first used as inputs in the bioreactor. Specifically, two geometries, one with a
confluence with the scaffold and another with a gap, were used. The next variable altered
was the porosity of the scaffold from the relative values determined with SEM and
ImageJ. The simulation then displayed various flow profiles through the bioreactor and
were coupled with a stress-strain analysis of the scaffold, and the von Mises stresses
Design criteria were met if the bioreactor and corresponding scaffold displayed
uniform flow profiles and consistent von Mises stresses throughout the scaffold volume.
This criterion should remain true for all relative porosity values, indicating that the same
55
Once the geometry was determined, multiple flow rates were used in order to
determine the flow rate or pressure value necessary to ensure von Mises stress of 5, 10,
and 20 dyne/cm2 throughout the volume of the scaffold. Analysis was simply performed
through an iterative design process until the optimal flow rates were determined.
Again, the final geometry was the key component of this analysis. No physical
bioreactor was fabricated in this study. The purpose was simply to optimize the system
and demonstrate that FEA can be used as an excellent design tool in bioreactor
Note: COMSOL model reports detailing the inputs and outputs from the
Statistics:
All numerical data were analyzed with one-way analysis of variance (ANOVA).
Significance was then determined with the Tukey Post Hoc Test. These methods assumed
populations, and Simple Random Samples (SRS). Furthermore, values were considered
56
RESULTS
2D Substrate Characterization:
dried and subsequently characterized with atomic force microscopy (AFM). AFM images
Figure 18. AFM Images of PS/PBrS 40/60 w/w polymer demixed substrata.
The figure above depicts PS/PBrS 40/60 w/w demixed substrata from
concentrations of 0.5%, 1.0%, and 2.0% after dissolution in chloroform. The 0.5%
nm in height. Note that the diameters of these nanoislands are random, and the
distribution of the surface topography is also random. Compare these substrata with those
nanoislands are 0.5 to 0.9 μm, 38 nm nanoislands with diameters of 0.7 to 1.0 μm, and 85
57
on average. This variation could thus correspond to a difference in overall surface area
also fabricated, dried, and analyzed with AFM. The AFM images from these analyses are
depicted in Figure 19 below. Titles listed above each figure are average island heights.
Figure 19. AFM Images of PLLA/PS 70/30 w/w polymer demixed substrata.
The image above depicts the substrata fabricated from 0.5%, 1.0%, 2.0%, and
3.0% PLLA/PS 70/30 w/w polymer demixing. Note that the corresponding nanoisland
heights were on the same order as the PS/PBrS 40/60 w/w demixed films; however, the
topographical characterization was much different than the other polymers. The
nanoscale for 0.5% films was approximately 12 nm in height, 1.0% films were around 21
nm in height, 2.0% films were approximately 45 nm, and the 3.0% demixed substrata
were around 80 nm in height. These substrata were even less uniform than their PS/PBrS
counterparts. The diameters also significantly changed from the 12-21 nm nanoislands to
the 45-80 nm nanoislands. Diameters for the 12-21 nm nanoislands were approximately
0.1-0.3 μm, and the diameters for the 45-80 nm nanoislands were approximately 0.5-1.0
58
μm. As before, this variation could hypothetically correspond to a variation in the surface
59
3D Scaffold Characterization:
20.
its calcium phosphate counterpart. The salt-leached PLLA scaffolds are shown in Figures
20.(E-J). Figures 20.E and 20.F depict the scaffolds with leached NaCl crystals ranging in
diameter from 150 to 300 μm at two different scales, 20.G and 20.H depict those from
leached salt crystals of 300 to 500 μm, and 20.I and 20.J depict PLLA scaffolds whose
leached crystals ranged in diameter from 500 to 710 μm. Note that the pore sizes
increased with the size of the NaCl crystals, but not by much. Furthermore, it appeared
that “pits” were actually leached from the PLLA polymer as opposed to true pores. It
would thus be hypothesized that proper cell seeding would not be possible in such
Using these SEM images, the relative porosities were then determined with
Figure 21. Estimated porosities calculated in ImageJ from SEM images of 3D scaffolds. (*p<0.05 w/ PLLA(300-500);
**p<0.01 w/ PLLA(300-500); ***p<0.001 w/ PLLA(300-500); +++p<0.001 w/ PLLA(500-710) N=10)
61
The “gel” as shown in this graph was another scaffold fabricated from ground
calcium phosphate, fibrin, and thrombin. The process was similar to the formation of a
blood clot in vivo. However, these scaffolds were not properly analyzed due to a
Note that the salt-leached PLLA scaffolds with 300-500 μm diameters displayed
porosities significantly greater than the BD© calcium phosphate scaffolds and the
BoneMedik© coral scaffolds. Furthermore, the PLLA scaffolds with NaCl crystals
than all other scaffolds, including the salt-leached PLLA scaffolds with 150-300 μm
porosities between the numerous scaffolds. However, note that the morphology of the
scaffolds shown previously have demonstrated that the coral and calcium phosphate
scaffolds actually have true, continuous pores when compared with the PLLA scaffolds.
Thus, it could be argued that these porosity estimates are not true indicators of cell
performance on each of the three dimensional scaffolds. Instead, these data will be
62
Mechanosensitivity of Stem Cells on Substrates:
Medium (DMEM) with 10% fetal bovine serum (FBS), 1% penicillin-streptomycin, and
1% L-glutamine. Randomly distributed nanoisland textures with varying heights (12, 21,
45, and 80 nm) were fabricated using poly(L-lactic acid)/polystyrene (70/30 w/w)
demixing techniques in concentrations of 0.5%, 1%, 2%, and 3% w/w. The solutions
were spin-casted on quartz slides at 4000 rpm for 30 s, completing substrate fabrication.
calcium ([Ca2+]i). Increased [Ca2+]i activates MAP kinases ERK1/2, resulting in increased
fluid flow chamber. Shear stresses of 5, 10, and 20 dyne/cm2 were induced by this
oscillating fluid flow. Fura Red AM stain was used to measure [Ca2+]i over time within
the cells through fluorescence microscopy. The initial 180s were static, and the final 180s
Figure 22 depicts the percentage of cells responding under static and fluid flow
conditions at each of the shear stresses after culture on each of the nanotopographies.
63
Figure 22. % Cell Response on various nanotopographies under different shear stresses
greater [Ca2+]i response compared with the flat control. Furthermore, shear stresses
dyne/cm2.
Figure 23. Absolute increase in fluorescence under fluid flow. (*p<0.05 w/ Flat;
64
Cells on 12 nm nanoisland topographies demonstrated significantly greater
Additionally, significant differences occurred between the 45-80 nm nanoislands and the
Finally, Figure 24 depicts the percent increase in fluorescence from a baseline set
Figure 24. Percent increase in fluorescence under fluid flow. (*p<0.05 w/ Flat; #p<0.05
w/ 12 nm; N=3)
under shear stresses of 5 dyne/cm2 and 20 dyne/cm2. However, the cells cultured on 80
displayed a significantly lower percent increase when compared with both the flat PLLA
65
FACS Analysis:
Multiple stem cell markers were used to determine whether substrate nanoscale
displayed randomly distributed nanoisland textures with varying island heights (11, 38,
and 85 nm). After annealing, PS segments segregate to the film surface, and hBMSCs
responses to nanotopographic scale could be assessed under the same surface chemistry
lower SSEA-4, CD73, and CD105 positive cell percentage relative to flat control and
larger nanoisland surfaces (Figure 26). After 12 days of culture, cells cultured in
growth media, regardless of textured or flat surfaces (Figure 27). Cell response to CD90
after 7 days of culture displayed similar trends but did not demonstrate significant results.
Control Control
11 nm 11 nm
38 nm 38 nm
85 nm 85 nm
Flat Flat
Figure 25. FACS Output for SSEA‐4 fluorescent markers.
66
The curves in Figure 25 display the fluorescence of a total of 10,000 counts.
Greater fluorescence would be shifted to the right. In each experiment, the fluorescent
counts were compared with the same control. A threshold was set for each experiment
based upon the control’s fluorescence level, and the percentage of counts beyond the
threshold (minus the artifacts in the control) allowed for a comparison in the percentage
of cells responding to each marker utilizing a FACS analysis. These data are displayed in
Figures 26/27. hBMSC response to SSEA-4, CD73, CD90, and CD 105 in growth and osteogenic media after 7 and 12
days of culture. (*p<0.05 w/ Flat; **p<0.01 w/ Flat; ##p<0.01 w/ 85 nm; N=3)
67
The above data depict the responses of hBMSCs cultured on 11 nm, 38 nm, and
Cells were cultured for either 7 or 12 days in growth or osteogenic differentiation media
and then tagged with either SSEA-4, CD73, CD90, or CD105 antibodies. Outputs similar
to the data shown previously were collected for each of these experiments, and the
First, note the general differences between all cells cultured in growth versus
osteogenic media on both Day 7 and Day 12. The general trend was a decrease in percent
Such a decrease was an expected result, and it demonstrated success in data collection.
Next, look at the data for cells cultured for 7 days in growth media. There were no
significant differences among the substrates for SSEA-4, CD90, or CD105 antibodies.
response to CD73 when compared with both the flat control and those cultured on 85 nm
nanoislands. Though this was significant, the response was still greater than 80%. Thus,
different results. First, note that significant differentiation of cells tagged with CD90
antibody diminished any significance among the substrata. However, cells cultured on 11
antibodies when compared with both the flat control and 85 nm nanoislands. This
68
differentiation of cells cultured on the 11 nm nanoislands. Again, this agrees with the data
when compared with other scales of topographies and flat controls. Again, surface
Focusing now on the cells cultured for 12 days in growth media, significant
results have also occurred. First, note that the overall cell responses to different
antibodies in growth media after 12 days were generally lower than those cultured for 7
days. Such differences could have been due to some natural differentiation occurring as
days in growth media displayed a significantly greater response to the CD90 antibody
when compared with the flat control. Furthermore, progenitor cells cultured on the 11 nm
nanoislands for 12 days in growth media also displayed a significantly greater fluorescent
response to SSEA-4 antibody when compared with the flat control. These increases could
substrata. However, this hypothesis conflicts with the response of cells cultured on 11 nm
demonstrated a generally lower response to all antibodies when compared with those
cultured in growth media. However, cells cultured on 11 nm nanoislands and tagged with
the flat control. These data are again correlated to an increased osteogenic potential, and
69
these data are once again coupled with the increased mechanosensitivity of stem cells on
12 nm nanoisland topographies. All other cells tagged with SSEA-4, CD73, and CD90
antibodies displayed such a low response that no significant differences were noted
70
AP Activity:
Human mesenchymal stem cells (hMSCs) were cultured on the BD© calcium
nanoislands topographies fabricated from PS/PBrS 40/60 w/w polymer demixing. The
progenitor cells were cultured in osteogenic differentiation media for 12 days. Cells were
lysed, and an alkaline phosphatase activity assay was performed for each of the scaffolds
and substrates (N=6). The AP activity was normalized by the number of proteins in the
***
Figure 28. AP Activity for cells cultured on flat PS and 11 nm nanoislands. (***p<0.001
w/ 11 nm; N=6)
significantly greater AP activity when compared with the flat substrate. These data are
further coupled with the data stating that progenitor cells cultured on 11-12 nm
71
nanoislands display significantly increased mechanosensitivity, osteogenic potential, and
now AP activity when compared with flat controls of the same surface chemistry (i.e.
PLLA or PS).
Depicted in Figure 29 below is the AP activity for each of the three dimensional
**
Figure 29. AP Activity for cells cultured on BD© calcium phosphate scaffolds and
significantly greater alkaline phosphatase activity when compared with the BD© calcium
phosphate scaffolds. In one sense, such a result is unexpected, due to the fact that the
BD© scaffold is currently undergoing FDA approval, but the BoneMedik© scaffold is
not. However, the analysis of the morphology of these scaffolds demonstrated previously
that cell seeding may be more efficient on the coral scaffold than on the calcium
phosphate scaffold due to increased continuity of more continuous pores on the coral
72
scaffold. Furthermore, the relative porosity of the BoneMedik© scaffold was also greater
73
Finite Element Analysis of Cell Confluence:
Numerous studies thus far have involved oscillatory fluid flow to induce shear
stresses on the surfaces of progenitor cells. However, the microscale and macroscale
patterns of flow have yet to be analyzed in detail. Finite element analysis (FEA) was
therefore employed to simulate fluid flow over these cells. In doing so, the data can then
predict both flow patterns and possibly cell responses to such flow. Cell biologists
typically do not employ such methods, making this particular study novel in its bio-
computational aspects.
unidirectional flow of media was employed. The pressure difference, as opposed to being
difference across the bioreactor, ρ is the density of the fluid in kg/m3 (in this case, water
was used with ρ = 1000 kg/m3), g is the acceleration due to gravity (g = 9.81 m/s2), and L
is the length of the bioreactor being simulated in meters. Figure 30 below depicts the
macroscopic view of flow over cells cultured on less hydrophilic surfaces with varying
levels of confluence. Note that there were very little differences in flow patterns within
the bioreactor.
74
Figure 30. FEA Analysis of cell confluence; macroscopic view. Incompressible Navier-
Although there were minor disturbances in the flow profiles, most notably with
50% and 100% confluence at the upper boundary of the bioreactor, significant
differences in flow patterns were not noted. Thus, it was recommended that a
75
microscopic view of flow over cells be analyzed due to the inconclusive results from this
preliminary study.
Figure 31 below depicts the microscopic view of flow over cells with various
confluence. Note the velocity field contour lines in addition to the color surface.
Figure 31. FEA of cell confluence; microscopic view. (A) 50% confluence with cell at
entrance to flow field; (B) 100% confluence with cell at entrance to flow field; (C) 100%
confluence with high hydrophobicity; (D) 50% confluence without cell at entrance.
Figures 31.A and 31.B depict unidirectional fluid flow of cells with 50%
confluence and 100% confluence, respectively. In both cases, a cell was located at the
entrance to the flow region. In doing so, this initial bump in the flow significantly altered
the flow patterns, such that flow was at or near zero along the surfaces of following cells.
Thus, it could be argued that the shear stresses along the surfaces of these cells would
effects, Figure 31.C depicts the flow along cells in such a bioreactor at the microscopic
76
level. Note that the flow is nonzero along the upper region of the cells, but the space
between cells exhibited flow rates at or near zero. Thus, this unidirectional flow
simulation suggests that the shear stresses between cells in extremely hydrophobic
conditions are also at or near zero. Furthermore, the shear stresses along the surfaces of
Finally, Figure 31.D depicts the culture of cells on hydrophilic surfaces at 50%
confluence beyond the entrance region. In this case, the lack of flow initially witnessed in
Figures 31.A and 31.B becomes minimal. Furthermore, some flow occurs over a greater
surface area of the cells, and the shear stresses become more uniform when compared
with those in Figure 31.C. Therefore, these data suggest that cells with a lower
confluence and on hydrophilic surfaces will have more uniform shear stresses under
77
Finite Element Analysis of Cell Height:
The next study attempted to examine the various effects of culture of cells on
surfaces differing in hydrophobicity under fluid flow conditions, using cell height as the
independent variable. Though the core of this thesis analyzes surface morphology under
constant surface chemistry, such chemistry should still be considered when developing a
substrata, it would be predicted that the apparent shear stresses would also vary as surface
chemistry is altered if cell height were to increase with hydrophobicity. First, consider the
extreme case of a cell with a very high contact angle with respect to the culture surface
Figure 32. Cell with a very high contact angle with respect to the surface under constant,
unidirectional flow.
78
This study demonstrates a few important characteristics of such extreme
conditions. First, the high cell height may lead to a lack of flow at edge of the cell on
both sides of flow. Such height change leads to a concentration of higher shear stresses at
the upper surface of the cell. It would thus be predicted that cell shear stresses with
increased cell height would be non-uniform in nature. As a result, oscillatory fluid flow
would be more beneficial than unidirectional flow in distributing the shear stresses along
Furthermore, note that eddies begin to form in the distal region of flow, just
beyond the right side of the cell in the flow field. These eddies can lead to turbulence if
velocity is increased, and such turbulence could hypothetically fabricate negative effects
As a result, this extreme condition predicts that oscillatory fluid flow provides a
unidirectional constant or unidirectional peristaltic fluid flow of media over cells extreme
The study then continues by examining oscillatory fluid flow and altering the cell
height among the simulations. The purpose was to determine the general differences
between relative shear stresses induced by oscillating fluid flow on cells cultured on
substrata with varying levels of hydrophobicity. It was hypothesized that cells with lower
height would exhibit greater shear stresses, but those with greater cell height would have
shear stresses concentrated near the center of the cell. This hypothesis stemmed from the
79
preliminary simulation shown in Figure 32. The results of these simulations are displayed
in Figure 33.
Figure 33. Shear stresses along the surfaces of cells cultured on substrata with varying
cell height. Oscillatory fluid flow of media was used, and these data depict the simulation
at t = 3.0 seconds. Red arrows depict the velocity field, and the grayscale gradient
depicts the von Mises stress distribution in the various simulations.
80
Note that the flow profiles were similar among the various conditions, except at
the cell surfaces. As cell height was increased, the fluid flow profile was further
disturbed. It is very difficult to quantify the von Mises stresses from these simulation
outputs, but the purpose is to depict and define the various levels of cell height involved
in this protocol. Figure 34 demonstrates the first quantification of these data in terms of
Figure 34. Shear stresses at the cell center over 3.0 seconds, along with the
First, note the wide variation in von Mises stresses from t = 0.0 s to t = 3.0 s
between cells cultured on various substrata. As can easily be seen, the amplitudes of these
sinusoidal patterns vary greatly, with those for the 7.5 µm-high cells having the greatest
peak amplitudes at nearly 120 dyn/cm2. The flat control was generally the lowest in its
However, upon averaging these data, the hypothesized trend that greater shear
stresses would be concentrated at the cell center on cells of greater height was nearly
81
observed. The flat control expectedly exhibited the lowest average shear stress at the cell
center, followed by the 1.5 µm- and 3 µm-high cells. Additionally, the very 7.5 µm-high
cells exhibited the greatest average shear stress at this location. However, the 5 µm-high
cells did not follow this trend. Such results could have been due to an error in the
simulation or some unpredicted factor in shear stress versus cell height. Nonetheless, all
Figure 35 depicts the same data for various substrata at the left adhesion point
between the cell and the surface under oscillating fluid flow for 3.0 seconds.
Figure 35. Shear stresses at the leftmost point on the cell over 3.0 seconds, along with
Unlike the previous location, the von Mises stresses at the leftmost point on each
cell from t = 0.0 s to t = 3.0 s was very similar in its sinusoidal pattern among cells
cultured on various substrata. In fact, it becomes very difficult to analyze these time-
dependent data at this location. Nevertheless, one can easily locate very high peaks for
82
Upon averaging these time-dependent shear stresses, one finds a trend very much
unlike that found at the center of the cell. The flat control exhibited the greatest average
shear stress over this period of 3.0 seconds. Furthermore, the trend demonstrated that
shear stress at this adhesion point decreased with increasing cell height. This trend was
hypothesized from the preliminary cell height data, predicting that most shear stresses
would be concentrated away from this point under conditions where height is greatest.
Finally, these data were again analyzed at the rightmost point on each cell over a
Figure 36. Shear stresses at the rightmost point on the cell over 3.0 seconds, along with
Considering that oscillatory fluid flow was used, similar trends in data were
displayed at the distal location in the flow field. These data further demonstrate the
benefit of oscillatory fluid flow over unidirectional flow, where such consistency would
not be exhibited as demonstrated previously for the cell with increased height. The trend
83
depicting decreasing shear stress with increasing cell height was once again noted at this
location.
Thus, these data demonstrate the benefits of decreased cell height and the use of
oscillatory fluid flow in the distribution of more uniform shear stresses across the surface
of cells.
84
FEA of Cells Cultured on Various Nanotopographies:
element analysis with COMSOL Multiphysics was employed to analyze oscillatory fluid
flow over cells cultured on nanoscale substrates. These substrates included plasma-
topographies. According to past analyses of these substrates [57], cells cultured on these
The purpose of these experiments was thus to analyze the differences in apparent
von Mises stresses under the same flow conditions. In order to do so, all cells were
fluid flow conditions. Additionally, the velocity fields were analyzed in order to
determine whether these conditions were upheld. As shown previously, velocity fields
should not be significantly altered when flow occurs over various nanoscale substrata
(Figure 30). Furthermore, hydrophilic surfaces and oscillatory fluid flow have been
demonstrated to exhibit the greatest level of shear stress distribution according to prior
Although it would be predicted that these assumptions will hold true, it is still
helpful to verify these assumptions by initially analyzing the velocity fields and von
85
Mises stress gradients at various time points and on the various substrata. These data
Figure 37. FEA simulation of oscillatory fluid flow on various substrata at t = 0.5 s, t =
2.3 s, and t = 3.0 s. Red arrows depict the velocity field, and the grayscale gradient
Though the size of the image makes it difficult to completely analyze these
simulations in detail, it is helpful to analyze each of the columns to compare among the
86
four substrata (Figure 37.a-d) at each time point. Notice that there are no significant
differences in the velocity field as demonstrated by the grayscale streamlines and the red
arrows. The middle column, at t = 2.3 seconds, depicts the peak velocity in the sinusoidal
fluid flow simulation. Again, even at the peak flow, there is no significant difference in
velocity field and von Mises stress distribution among the various surfaces. This analysis
allows for a consistent examination of the apparent shear stresses along the cell surface.
In order to analyze the average shear stresses along the cell surface, the von Mises
stress was quantified for each element along the arc length of the cell surface. Each of
these elements was then averaged with the others to find this apparent shear stress at the
cell surface. These average shear stresses at t = 0.5 s, t = 2.3 s, and t = 3.0 s are depicted
in Figure 38 below.
Figure 38. Average shear stresses in the FEA of oscillatory fluid flow over various
shear stresses than both the cells cultured on plasma-cleaned glass and those cells
87
significantly greater in apparent shear stresses than those cultured on flat polystyrene.
Thus, the apparent shear stresses exhibited under oscillatory fluid flow was always
significantly greater on the various nanoscale topographies than their flat counterparts at
significantly greater than those cultured on plasma-cleaned glass and flat polystyrene.
These data suggest that the significant increase in apparent shear stresses is upheld under
Finally, at 3.0 s, the significance in data similar to the t = 0.5 s time point
greater in apparent shear stresses than their flat counterparts. Furthermore, those cultured
on 11 nm nanoislands exhibited significantly greater apparent shear stresses than the cells
The key result of this study was the finding that cells cultured on 11 and 38 nm
nanoisland topographies exhibited significantly greater apparent shear stresses under the
same oscillatory flow conditions than their flat counterparts. This increase in apparent
cells cultured on these nanoscale substrata when compared with flat controls.
88
3D Bioreactor Design by the Finite Element Method:
properties. The design of such a two-dimensional bioreactor has been analyzed above, but
is induced must now be developed with the five design criteria set forth in the project
The bioreactor design must satisfy the design criteria by (1) maintaining a
uniform shear stress throughout the scaffold volume, (2) ensuring the bioreactor diameter
is between 2 and 10 mm, (3) allowing for 1 Hz oscillations of fluid flow over time
through a symmetrical geometry, (4) ensuring all velocities are greater than zero in the
scaffold volume, and (5) repeating the optimal geometry simulation for each previously
analyzed scaffold by comparing simulations with the relative porosities calculated earlier.
10 mm diameter requirement, were first created with COMSOL Multiphysics. The first
geometry suspended the scaffold in the center of the bioreactor, and the second geometry
tightly held the scaffold against the bioreactor walls. First, however, the general geometry
89
Scaffold
Figure 39. Initial geometry with FEA setup for analysis of scaffolds. This geometry has a
diameter between 2 and 10 mm, and it is symmetrical for use with 1 Hz oscillations of
fluid flow.
The upper-left image in Figure 39 depicts the general geometry. Note that it is
symmetrical for use with 1 Hz oscillatory fluid flow. The upper-right image depicts the
volume scale factor. As can be seen, fluid flow will occur more easily throughout the
length of the bioreactor, but shear stresses will be very high where the volume scale
factor is very low (at the two flow ports and the center of the bioreactor). This
significantly decreases required computing power for FEA. The bottom-right image then
90
depicts the area of interest, where the scaffold will be placed. The numerical values in
this design were arbitrary at this point in order to show general geometry and design
setup.
The next phase of the design process then compared the tight and loose scaffold
geometries. The purpose was to determine which geometry would better satisfy Criterion
1, which states that the bioreactor must exhibit uniform shear stresses throughout the
volume of the scaffold. A simulation of the BD© scaffold using the Brinkman model for
porous media in both tight and loose geometries is shown in Figure 40.
Figure 40. Bioreactor simulations of the BD© scaffold in tight and loose geometries.
Figures 40.A and 40.B compare the stresses between both loose and tight
geometries for the BD© scaffold as shown by the color gradient. Note that the tight
91
geometry (Figure 40.B) has a high concentration of shear stress near the axis of
symmetry, with very little von Mises stresses near the outer perimeter of the scaffold. On
the other hand, the loose geometry exhibited von Mises stresses with a more uniform
distribution throughout the volume of the scaffold. The uniformity was far from perfect,
but this novel design demonstrates a benefit to loose-fitting bioreactors with their
scaffolds.
Figures 40.C and 40.D compare the velocity fields between the loose and tight
geometries for the BD© scaffold. In both cases, most of the flow was concentrated near
the center of the scaffold with very little flow near the outer perimeter. Thus, perfusion of
media throughout the scaffolds followed nearly the same pattern between the two
geometries.
Therefore, the loose geometry was considered the better choice due to its
satisfaction of Criterion 1. The shear stresses were more uniformly distributed throughout
the volume of the BD© calcium phosphate scaffold in this particular geometry, providing
nearly the same between the two bioreactor geometries, so the key factor was simply the
With the optimal bioreactor geometry chosen, the next phase was to ensure that
velocity throughout the volume of the scaffold remained greater than zero for each of the
aforementioned scaffolds. To do so, each scaffold was simulated in the loose geometry
bioreactor, and the velocity field was quantified. The velocity of each element along the
arc length of the scaffold from the axis of symmetry (r = 0) to the outer edge of the
scaffold (r = R) was determined, and these were than averaged along the z-axis of the
92
scaffold. Provided was an average velocity gradient from r = 0 to r = R for each scaffold
under the same conditions. The output of these analyses is depicted in Figure 41 below.
Figure 41. Average velocities from r = 0 to r = R in the loose bioreactor geometry for
each of the scaffolds. These data were then averaged again along the arc length to
First, note that the prediction of a greater velocity near the center (r = 0) and a
diminishing velocity near the outer perimeter (r = R) was verified according to the graph
on the left in Figure 41. Furthermore, Criterion 4 was satisfied by demonstrating that the
velocity did not drop below zero on average throughout the arc length of each scaffold. It
is very interesting to note that the velocity gradients appear very similar among all the
scaffolds. Upon averaging these data along the arc length of the scaffold, the graph on the
right in Figure 41 further demonstrates that these velocities do not significantly change as
porosity is significantly altered (Figure 21). Thus, this same bioreactor geometry could be
The final criterion, Criterion 5, required that all the scaffolds be analyzed and
their shear stresses (in this case, von Mises stresses) be compared. The purpose is to
ensure that the same bioreactor could hypothetically be used for numerous scaffolds
93
under the same protocol. Using the same method as in Figure 41, von Mises stresses were
now calculated along the arc length of each scaffold in the loose bioreactor geometry
Figure 42. FEA simulation of von Mises stress in the loose bioreactor geometry for each
Note once again that all of the scaffolds displayed very similar distribution of
shear stresses along the arc length of the scaffolds from r = 0 to r = R. These data were
then averaged once again, and the graph on the right of Figure 42 depicts the average
difference in average shear stress among the various scaffolds. However, the patterns
mimicked the porosities of each scaffold (Figure 21). Nonetheless, this lack of
significance demonstrates that the bioreactor could be used for numerous scaffolds with
significantly different porosities under the same protocol. As a result, Criterion 5 was
thus satisfied.
94
Thus, the results come full-circle from the simplification of a three-dimensional in vivo
upon such prior data regarding the benefits and analyses of two-dimensional culture in
bioreactors in vitro.
95
DISCUSSION
Two-Dimensional Substrata Characterization:
magnitude equivalent to nanometer [1, 48]. The images from atomic force microscopy
By increasing the concentration of polymer in chloroform solvent, the nanoscale was thus
increased. PS/PBrS 40/60 w/w films were found to elicit nanoislands increasing from 11
nm to 85 nm, while the solute (i.e. polymer) concentration was increased from 0.5% to
the same order of magnitude [13]. Though the number varies, it is widely believed that
that these polymer demixing techniques successfully mimicked the nanoscale of in vivo
bone tissue.
with such a method is that it does not successfully mimic the in vivo environment, due to
the fact that the extracellular milieu of bone is in fact very disordered. In contrast with
lithographic methods which could already successfully create the same nanoscale as
96
previously mentioned, this study wished to provide a disordered patterning of
nanoislands. When analyzing the images from AFM, one can easily witness a random
assortment of nanoislands. The height varies greatly among the islands, along with the
the surface of the film. Such random distribution demonstrated success of these polymer
the cantilever. Even with a sharp cantilever, AFM images depict rounded surfaces where
surfaces are instead sharp. Thus, it should be noted that the round nature of these islands
Another issue results from the random character of the nanoscale topographies.
For example, larger (80-85 nm; Figures 18 and 19) topographies demonstrate an
islands actually vary in height to an extreme degree. Furthermore, the width of these
97
Three-Dimensional Scaffold Characterization:
various scaffolds were revealed. Firstly, the BD© calcium phosphate scaffolds (Figures
20A and 20B) demonstrated similar textures and topographies when compared with the
BoneMedik© coral scaffolds (Figures 20C and 20D). However, the calcium phosphate
scaffolds did not appear to have deep pores, and the scaffold instead illustrated a series of
construct. The coral scaffolds, on the other hand, demonstrated pores with much deeper
calcium phosphate scaffold images depicted much larger topographies when compared
texture than both the coral and calcium phosphate scaffolds. As the NaCl salt crystals
were increased in diameter, polymer “shells” became larger accordingly. However, the
pores in these scaffolds were not true pores; instead, the topographies demonstrated
microscale pits as opposed to true microscale pores. These images became integral in
understand the differences in cell and media perfusion among the constructs.
Textures depicted from SEM images of the scaffolds could then be compared with
constructs were illustrated with a rough texture similar to that of the scaffolds found in
the poly(L-lactic acid) constructs [5]. When compared with scaffolds created from
98
nanofibers, the coral and calcium phosphate scaffolds demonstrated the deep pores found
in these constructs [66, 71-73]. Such deep pores became useful in the perfusion of media
leading to efficient cell seeding. Thus, these data provided beneficial insights into the
drawbacks of collagen scaffolds through analyses of pore depth and construct texture and
The estimates of porosity were then calculated from these SEM images utilizing
increased estimated porosities in the polymer scaffolds. However, such findings must be
met with a skeptical attitude, especially due to the aforementioned lack of pore depth in
One of the greatest issues with the estimation of porosity is the concept that three-
representation severely loses its accuracy. A mock scaffold is depicted in Figure 43.
A B
Figure 43. A comparison of the 2D representation of a 3D construct (A) with the actual
3D construct. Notice the loss of depicted pores in these representations.
99
In Figure 43, notice that the two-dimensional image in Figure 43A depicts only
five pores. ImageJ would then be used to calculate the porosity, estimating that these five
pores are continuous throughout the volume of the construct. Thus, it would be estimated
Figure 43B depicts pores lost in the two-dimensional image. As can be imagined, the
actual porosity may thus be much greater than the estimate due to the loss of pores by
100
Stem Cell Growth on 2D Substrates:
(hMSCs) cultured on PLLA/PS 70/30 w/w demixed films was studied by analyzing the
relative fluorescence of Fura Red stain. When analyzing the percentage of cells
responding in Figure 22, one finds that, at a shear stress of 5 dyne/cm2, hMSCs cultured
PLLA controls. At greater stresses (10 and 20 dyne/cm2), the responses were too great to
determine any significance. Such data illustrate the observation that mechanosensitivity
the absolute fluorescent response above the average baseline during static conditions. It
was observed that cells cultured on larger (45 nm and 80 nm) nanoislands displayed
the mechanosensitivity. Thus, it was observed that cells not only preferred nanoscale
topographies, but a preference existed for certain scales (i.e. 12 nm) over others (45 and
80 nm).
Finally, Figure 25 depicted the percent increase in fluorescence above the baseline
during static conditions. Similar to the data in Figure 24, hMSCs were observed to exhibit
some preference among the substrates. These data thus became critical in the
101
development of a bioreactor. If cells exhibit preferences in mechanosensitivity at
different fluid flow induced shear stresses and also among varying nanoscale
bioreactor design used oscillatory fluid flow, along with varying three-dimensional
Utilizing data from fluorescence activated cell sorting (FACS) analysis (Figure
marrow stromal cells (hBMSCs) cultured on PS/PBrS 40/60 w/w demixed films.
Specifically, these films displayed nanoscales of 11 nm, 38 nm, and 85 nm. A flat
polystyrene control was also analyzed. hBMSCs were cultured on these substrates for
either 7 or 12 days, and harvested cells were tagged with SSEA-4, CD73, CD90, or
After seven days, the cells were harvested from culture in either growth or
media displayed significantly lower SSEA-4, CD73, and CD105 reactivity compared
with the flat control and 85 nm nanoisland substrates. This decrease in reactivity was
nm nanoislands.
After twelve days, the cells were harvested from culture in either growth or
significantly greater reactivity to SSEA-4 compared with the flat control. Such reactivity
102
was directly correlated with an increase in proliferation potential for cells cultured on
displayed significantly lower CD105 reactivity when compared with flat controls. These
data further verified the observation that cell differentiation potential was increased on
these substrata.
counterparts. Combined with previous data, we can now infer that substrate topography
Previous studies have also verified that focal adhesions also increase on the same
substrata [1]. Combining all of these data allows for the observation that nanoscale
PS/PBrS 40/60 w/w demixed films and flat polystyrene substrata were harvested after 12
After comparison of the two samples, it was observed that cells cultured on 11 nm
hMSCs cultured on flat polystyrene controls. Thus, cells have been demonstrated to
103
When these data are compared with previous studies, the data demonstrate a
potential, and differentiation potential. Thus, surface morphology alone can regulate stem
cell activity.
104
Stem Cell Growth on 3D Scaffolds:
substrates, human mesenchymal stem cells (hMSCs) were also studied on three-
BoneMedik© coral scaffolds for 12 days in osteogenic media prior to harvesting. These
samples were then analyzed for alkaline phosphatase (AP) activity, normalized by the
After analysis of these data, it was observed that progenitor cells cultured on the
BoneMedik© coral scaffolds displayed significantly greater AP activity than those cells
This observation was expected when considering the differences in porosities and
surface morphologies (Figures 20, 21). The coral scaffold was observed to have a greater
porosity value than the calcium phosphate scaffold. Additionally, the morphology of the
coral scaffolds were observed to have deeper and more uniform pores than the BD©
Thus, it could be argued that increased perfusion and deeper pores allowed for better flow
of osteogenic media throughout the volume of the scaffolds. Such flow, in turn, would
105
Comparison with 2D Substrata:
One will notice that the alkaline phosphatase activities were not directly
samples.
However, the number of sigma units per protein was on average greater among
Specifically, averages for two-dimensional substrates ranged from 125-250 sigma units
per protein, whereas those for three-dimensional scaffolds ranged from 10-70 sigma units
per protein. When comparing the calcium phosphate scaffolds with the two-dimensional
These observations should be taken with a grain of salt, due the aforementioned
differences in protocol and lack of statistical analyses between two- and three-
106
Finite Element Analysis:
FEM simulations were used in this study in order to analyze fluid flow patterns
inside a cell cultured flow chamber. Firstly, we examined the variability of fluid flow
patterns among various levels of cell confluence. At the macroscopic level, no significant
differences were noted. Thus, cell confluence was concluded to have no major effect on
the flow profiles in bioreactors. However, microscopic analyses demonstrated that little
to no flow would occur between highly confluent cells. When cell confluence was
decreased, fluid flow could more easily occur between cells. Furthermore, cells in the
entrance region of flow in the bioreactor would alter the patterns of flow over cells
downstream. It would thus be assumed that lower levels of confluence, 50% confluent or
less, would be optimal in the design of bioreactors utilizing such fluid flow. This lower
confluence allows for a better distribution of fluid flow across cell surfaces and thus a
more even distribution of wall shear stresses. However, we suspect that the lower
communication. Thus, there may be an optimal cell confluence level that allows both
As regards cell height effects, FEM demonstrated a lack of flow in the distal
region of cells when the cell height is greater. Furthermore, these data suggest that
oscillating fluid flow used in some bioreactor systems [8, 39, 40] would be more
bioreactors due to a more uniform distribution of shear stresses along cell surfaces [7, 35,
74, 76, 77, 81]. In analyses using oscillating fluid flow conditions, cells with larger
107
heights lead to more concentrated shear stresses near the center of the cell. With
decreasing cell height (relevant to increased cell spreading), the shear stresses at the
center of the cells decreased on average and the shear stresses at the left and right of the
cell increased. These relative changes imply a more efficient distribution of shear stresses
with decreasing cell height. Taken together, these data suggest that bioreactors should use
oscillating fluid flow in combination with substrates that stimulate cell spreading for a
moduli, cell adhesion, and morphology [1, 45, 57]. Continuing with finite element
analyses utilizing oscillating fluid flow, data from these studies was implemented to
determine any potential differences in the apparent shear stresses of cells cultured on
these various substrates. It was demonstrated in this study that cells cultured on 11 and 38
nm nanoislands exhibited a significantly greater apparent shear stress than cells on flat
surface. Such shear stress variation was due to an increased elastic modulus present in the
cells cultured on these surfaces. Increased cell stiffness is correlated with an increased
apparent shear stress, and these data support the hypothesis that cells with a greater
elastic modulus will exhibit increased apparent shear stresses under oscillating fluid flow
conditions.
108
Benefits over Empirical Data Collection:
factors not easily determined empirically through the demonstration that cells cultured on
flat surfaces. Furthermore, FEM was used in such a way that an experiment could be set
up to empirically validate these results through parallel analyses. Thus, these methods
A key element of the data presented was the demonstrated benefit of oscillating
fluid flow as a superior method to its unidirectional counterpart. Flow is more evenly
distributed in the bioreactor system over time, leading to a more even distribution of von
Mises stresses across cell membranes. This distribution allows such systems to be more
predictive due to a better estimate of the flow rates and shear stresses at the walls of a
bioreactor. Thus, it can be concluded that oscillating fluid flow provides greater benefits
109
Summary of FEM as a Tool in Cell Culture
FEM was successfully used to analyze flow patterns over cells, as demonstrated
by the data previously presented. FEM provides an excellent tool in the prediction of
experimental protocols and the design of bioreactor systems. Though the bioreactor
plane, FEM can easily be expanded to analyze more complex geometries. When
method could provide insight not easily acquired through empirical methods.
110
Bioreactor Design:
uniform shear stresses were witnessed in the loose variation of the bioreactor design, as
depicted in Figure 40.A. The shear stresses throughout this particular design’s volume
were uniform.
Additionally, it was specified that the design must satisfy physiologically correct
shear stresses. The target ranges were 5, 10, and 20 dyne/cm2. By altering input
velocities, the shear stresses could be adequately controlled. Average shear stresses were
shown to be 5.36, 10.54, and 21.88 dyne/cm2 in this particular model. Figure 44 depicts
the changes in shear stress along the r=0 to r=R arc length of the scaffold.
Figure 44. Average shear stresses along the arc length of the loose scaffold model were
successfully maintained at approximately 21.88, 10.54, and 5.36 dyne/cm2.
111
Criterion/Specification 2: Standard scaffolds must fit within the bioreactor volume.
The bioreactor must have a variable diameter ranging somewhere between 2.0 and
10.0 mm.
The bioreactor is depicted in Figure 39. Though it may be difficult to witness, the
height of the scaffolds were set to either 3 mm (loose) or 4 mm (tight). In both cases, the
bioreactor was modeled such that it could support scaffolds ranging in diameter from 5
mm to 8 mm. This value was well within the range specified previously.
Though the bioreactors simulated in Figure 39 and Figure 40 did not undergo
oscillating fluid flow, the benefits of 1 Hz oscillations were previously determined in the
However, the symmetrical nature about the r-plane should lead to the capability of
Flow profiles were depicted in Figure 40 and Figure 41. Figure 40 demonstrated
through a surface map of both tight (Figure 40.D) and loose (Figure 40.C) scaffolds that
the loose bioreactor design led to a more uniform distribution of velocity through the
Additionally, velocities along the arc length of the loose scaffold was simulated
for different types of scaffolds. In all cases, a parabolic velocity profile was maintained,
112
where velocity was greatest at r=0 and smallest at r=R. Even with no-slip conditions
The loose bioreactor design (Figures 40.A and 40.C) was chosen as the optimal
bioreactor design. This bioreactor was simulated with various porosity values as
determined from SEM characterization of scaffolds. In doing so, the changes in shear
stress and velocity along the arc length of each scaffold was determined. Figure 41 and
Figure 42 depict these differences. It was observed that scaffold porosity changes did not
correlate to significant changes in velocity profiles or shear stress profiles along the arc
length of each scaffold. According to these FEM simulations, the bioreactor should
The tight bioreactor model was far less successful than the loose bioreactor model
in exhibiting uniform velocity profiles and shear stresses throughout the volumes of
depicted in Figure 40. Notice that the loose bioreactor is more capable of exhibiting
uniform shear stresses throughout scaffold volumes. Such stress distribution would thus
be correlated to an increase in the scaffold’s perfusion ratio compared with that of the
tight bioreactor.
The only issue with the loose bioreactor model is the need for secured suspension
of the scaffold within the bioreactor. This particular model assumed that the scaffold
113
simply floated in the volume of the bioreactor, which could potentially cause harm to
both the cells and scaffolds alike. To prevent this issue, it would be beneficial to include
some type of securing mechanism with adjustable supports one order of magnitude
smaller than the bioreactor diameter. This difference in size should prevent issues in
disruption of fluid flow. Empirical data must then be collected after fabrication of the
scaffold with the securing mechanism in place. Such a design is depicted in Figure 45.
Flow
r
Supports with
z
Screws for
Adjustment
Scaffold
Figure 45. Loose scaffold with six tight cables to suspend the scaffold within the volume
of the bioreactor.
Analysis of Design:
This bioreactor design satisfied all of the design criteria and specifications set
forth previously. Additionally, further design components have been prepared in order to
Before the design can be considered a success, however, perfusion studies must
114
unnidirectional or oscillatoory pump. Cells
C of anyy type will then
t be perfu
fused througgh the
The next
n phase would theen be to analyze
a the effects of oscillatory and
Specifically, the initial teest would bee an alkaline phosphatase assay afterr varying peeriods
of fluid flow and with diffferent shearr stresses. Thhis design scchematic is outlined
o in Figure
F
46.
Analysis oof
long‐term m
fluid flow of
Analysis off
A
progenito or
perfusion
cells in the
capabilitiees of
volume of
cells in varrious
various
Fabricattion of scaffolds wwith
scaffolds
bioreacttor t
the newly‐ ‐
within thiis
from CAAD f
fabricated
bioreacto or.
model. bioreactorr.
115
CONCLUSIONS
Methods of Biomaterial Characterization:
chemistry properties can be analyzed with X-ray photoelectron spectroscopy (XPS), and
tensiometry can be analyzed with contact angle measurements, for example. However,
the key analysis in this study was surface morphology. In order to do so, both atomic
force microscopy (AFM) and scanning electron microscopy (SEM) were involved.
discovered that AFM was very successful in predicting surface morphology. The method
not only provided useful data, but it was simple to employ. The significant drawback to
data from AFM was that it will oftentimes depict rough surfaces with rounded edges.
Such an artifact is due to the nature of the cantilever and was not corrected in the
scaffolds. It was very useful in estimating the pore sizes and thus the relative porosities of
numerous scaffolds. However, these relative porosities varied significantly from the true
porosity values, as illustrated previously. Thus, other methods than electron microscopy
would be beneficial to determine the total available void volume relative to the total
Therefore, the methods involved in this study were successful in that they
dimensional scaffolds. However, these morphologies had their flaws, though the most
significant flaws resulted from estimates of three-dimensional porosity values from the
116
tw
wo-dimensio
onal surfacee images inn the SEM characterization of thhe various three-
dimensional scaffolds.
It wou
uld be benefficial in futuure studies too further chaaracterize thhe morphologgy of
m
methods havee been illustrrated in Figuure 47.
•En
nergy Dispersiive X‐Ray Speectroscopy
•Fo
ocused Ion Beam
Miccroscopyy nning Optical Microscopy
•Neear‐Field Scan
•Traansmission Ellectron Micro
oscopy
•Coonfocal Raman Spectroscopy
Surfacce and TThin •X‐Ray Photoeleectron Spectro
oscopy
Film •Seecondary Ion M
•Au
Mass Spectro
oscopy
uger Electron Spectroscopyy
Strructural •Sm
mall Angle X‐RRay Scatteringg
Analysis •X‐Ray Diffractio
on
Optical
O •Ulttraviolet‐Visib
ble Spectrosccopy
Specctroscoppy •Fo
ourier Transfo orm Infrararedd Spectroscop
py
•Ind
ductively Cou
upled Plasma‐‐Atomic Emission
Spectroscopy
Chemical Analysis •Ind
ductively Cou
upled Plasma‐‐Mass Spectro
oscopy
•Ion
n Chromatography
•Po
olarization and d Strain Meassurement
Physiccal Propeerty •Co
ontact Angle M Measurement
Determination •Diffferential Scanning Calorim
•Acccelerated Surface Area an
metry
nd Porosimetrry
117
A
Ability of Su
ubstrates and Scaffolds to Regulatee Stem Cell Activity:
chhanges in po
orosity amonng scaffolds, were used as
a independeent variabless to determinne the
The previously prresented dataa have demoonstrated the utility of sppecific nanoisland
M
Mechanosens Ca2+]i) respponse,
sitivity, as determined through inntracellular calcium ([C
coompared witth flat surfacces and alsoo among the 12 nm nanooislands com
mpared with other
sccales. Simillar results also resultted from thhe data deescribing inncreased alkkaline
phhosphatase (AP) activity, increaseed stem celll proliferatioon, and increased stem
m cell
differentiation
n. The key was that thhese data illlustrated a preference
p f stem cells on
for
nanoscale su
ubstrates com
mpared withh flat surfaaces. Additioonally, speccificity arose for
sppecific (11-3
38 nm) nannoscale substtrata over thhose substraates of a larrger scale. These
T
coonclusions have
h been suummarized inn Figure 48.
Stem Cell Respo
onsiveness to
Two
o‐Dimenssional Surface Mo
orphologgy
surrface morphhology.
118
It should also be noted that increased responsiveness has been correlated with
previous studies of increased elastic moduli among cells cultured on these specific two-
dimensional nanoscale substrata. This relationship between cell stiffness and biological
also briefly analyzed. Unlike the studies of two-dimensional substrata, only alkaline
phosphatase activity was assayed. After analysis, the data illustrated that the
the BD© calcium phosphate scaffold. This variance could have been a result of
differences in porosity as previously discussed. However, the key finding was that not all
scaffolds are created equal. Whether this is a function of surface chemistry, surface,
could include studies of AP activity under oscillating fluid flow conditions, analyses of
fluorescence-activated cell sorting (FACS). These studies would thus parallel those
119
F
Finite Elemeent Method as a Novel Tool
T in Tisssue Engineeering:
H
However, flu
uid flow proffiles and thee affects of biophysical
b s
stimuli on cell surfaces were
orr some simillar macroscoopic parametter. Finite ellement analyyses, on the other
o hand, allow
a
inn vitro, and thus the in silico methhod providess a novel tecchnique to extract
e otherrwise
unnknown datta. When cooupling fluidd flow patterns of cells in culture with
w stress-sstrain
annalyses affo
orded by muultiphysics, or
o finite elem
ment analysiis, software,, the downsttream
efffects of these biophysiccal stimuli caan be prediccted. Previouusly illustrateed was the utility
u
of the software
r in predictting relative shear stressees (i.e. von Mises
M stressees) imposed upon
The other
o factor that can be
b useful is the use off finite elem
ment analyses in
shhear stressess at various locations aloong the surffaces of cellss. Such dataa cannot be easily
accquired emp
pirically, annd FEM thuus provides an excellennt alternativve. Some of the
Macroscopic
M c Functions Microsscopic Functtions
• Averrage flow proofiles • Flow alongg surfaces
• Averrage shear sttresses • Surface shear stresses
120
The applications in Figure 49 have been divided in macroscopic and microscopic
utilities. It should be noted, however, that these were simply the uses demonstrated in this
Another novel application of the finite element method in tissue engineering was
found in the design of bioreactors. FEM provides the ability to predict perfusion, flow
properties, deformation, stresses, and more with various geometries and inputs for
turbulence result in a mixing chamber. FEM can predict this through both velocity
profiles and Reynolds number calculations. The utility of FEM as a design tool in tissue
Thus, the finite element method acts as an excellent tool for predictive studies in
materials and bioreactors determined in vitro can be coupled with physical outputs
determined in silico.
121
Summary off Bioreactorr Properties Key for Su
uccess:
Loose/Seccure Scaffold Fitting
R‐ and Z‐symmetry Media must flow even arround the
Thiss allows for th
he utilization of scaffold, but the fluid flow should not
w during cell culture.
oscillating fluid flow aadversely affeect cell growth
h by jarring
the scaffo
old from its poosition.
Key Bioreacctor
Propertie
es
Laaminar Flow
Increased Perfusion
Turbulent flo
ow leads to lo
oss of cell
The optimal bioreacttor should maaximize
aadhesion, andd the Reynoldds number
perfusion of media a
o and cells throu
ughout
m
must thus be less than 230 00, without
the volu
ume of the sccaffold under study.
any eddiees in the flow
w field.
The parameters
p ouutlined in Fiigure 50 deppict those coommon among the bioreeactor
saatisfying all design criteeria. Symmeetry both raddially (r-direection) and longitudinall
l ly (z-
direction) is required
r for the utilizatioon of oscillatting fluid floow. Oscillatiing fluid flow
w has
m
more beneficcial than unnidirectionall flow. Thuus, the optiimal bioreactor would take
crriterion requ
uires that flow
w exist arouund all outer surfaces of the
t scaffold (i.e. loose) while
w
w the key
was y parameterr when com
mparing twoo major bioreactor geeometries. Those
T
122
bioreactors with a loose fitting illustrated better distribution of flow rates and shear
throughout the volume of the scaffold being studied. Increased perfusion correlates to
adhered cells, and thus more efficient tissue growth in the scaffold. As a result, the key
has been demonstrated to break down focal adhesions and shear cells off surfaces. Such
flow would lead to a degradation of tissue growth efficiency in the scaffold. If the
Reynolds number is kept below 2300 and eddies do not appear in the current, it can be
assumed that laminar flow is maintained in the device. Such a parameter must be
Bioreactor designs in the future should use FEM as a design tool. The utility of
this method in tissue engineering has been adequately demonstrated in this paper through
a preliminary design. The purpose of this study was to demonstrate key parameters for
bioreactors and introduce a new method for bioreactor design. With these goals in mind,
123
Closing Remarks:
alone involved skills from cell biology, bioengineering, computer science, and
biochemistry. Such a diverse field requires numerous approaches, and a new method has
dependent upon legislation over cell line restrictions, support through funding sources,
numerous abilities and limitations present in the field have hopefully been illuminated.
been analyzed in lieu of the design of an entire system. Both two-dimensional and three-
were then analyzed and compared. The field of biomaterials is yet another discipline to
be added to the list, one whose governing laws have not yet been confirmed. With so
techniques was required. In the end, only the cellular response could be analyzed. This
provided an optimal set of materials with specific parameters upon which cells will
The finite element method was introduced as a novel technique for tissue
required for bioreactor optimization were determined. It was determined that FEM would
124
As a result, a new approach to tissue engineering is proposed. This process is illustrated
in Figure 51.
Preliminary
Computational
Modeling (FEM, in
silico)
Assessment of Bioreactor
Bioreactor Success Fabrication (in
or Failure vitro)
Cell Culture in Stem Cell
Bioreactor (in vitro) Harvesting (in vivo)
Figure 51. An iterative design process for design of “in vitro” bioreactors, including a
As depicted in Figure 51, the process will begin the computational modeling of
stem cells will be harvested and cultured in the in vitro bioreactor. After the study is
complete, data will be analyzed to determine successes and/or failures in the design.
These will then be modeled by FEM once again for re-optimization, and the process will
continue. Such an approach deviates slightly from the traditional approach to skeletal
tissue engineering.
This study was thus successful in demonstrating the complete process as depicted
in Figure 51. All of the specific aims set forth in the Introduction have been
125
accomplished, and this study has thus been a success. The field of tissue engineering is a
vast terrain waiting to be explored. These pages have only delved into the outer perimeter
of the field, hopefully whetting one’s appetite for further exploration. It is the hope of this
investigator that such exploration will not only be considered but also pursued.
126
REFERENCES
1. Lim, J.Y., et al., Osteoblast adhesion on poly(L-lactic acid)/polystyrene demixed
thin film blends: effect of nanotopography, surface chemistry, and wettability.
Biomacromolecules, 2005. 6(6): p. 3319-27.
2. Wheeler, D.L. and W.F. Enneking, Allograft bone decreases in strength in vivo
over time. Clin Orthop Relat Res, 2005(435): p. 36-42.
3. Bancroft, G.N., V.I. Sikavitsas, and A.G. Mikos, Design of a flow perfusion
bioreactor system for bone tissue-engineering applications. Tissue Eng, 2003.
9(3): p. 549-54.
4. Benoit, D.S., S.D. Collins, and K.S. Anseth, Multifunctional hydrogels that
promote osteogenic hMSC differentiation through stimulation and sequestering of
BMP2. Adv Funct Mater, 2007. 17(13): p. 2085-2093.
5. Dawson, J.I., et al., Development of specific collagen scaffolds to support the
osteogenic and chondrogenic differentiation of human bone marrow stromal cells.
Biomaterials, 2008. 29(21): p. 3105-16.
6. Trojani, C., et al., Three-dimensional culture and differentiation of human
osteogenic cells in an injectable hydroxypropylmethylcellulose hydrogel.
Biomaterials, 2005. 26(27): p. 5509-17.
7. Zhao, F. and T. Ma, Perfusion bioreactor system for human mesenchymal stem
cell tissue engineering: dynamic cell seeding and construct development.
Biotechnol Bioeng, 2005. 91(4): p. 482-93.
8. Donahue, T.L., et al., Mechanosensitivity of bone cells to oscillating fluid flow
induced shear stress may be modulated by chemotransport. J Biomech, 2003.
36(9): p. 1363-71.
9. Felix Bronner, M.F.-C., Antonios Mikos (eds.), Engineering of functional skeletal
tissues. Vol. 1. 2007: London : Springer.
10. MedLine-Plus. Bone Grafts. [cited; Available from:
http://www.nlm.nih.gov/medlineplus/bonegrafts.html.
11. Endo, T., S.V. Bryant, and D.M. Gardiner, A stepwise model system for limb
regeneration. Dev Biol, 2004. 270(1): p. 135-45.
12. Ulrich Meyer, H.P.W., Bone and cartilage engineering [electronic resource]
2006, Berlin; New York: Springer.
13. Hollinger, J.O., Bone tissue engineering. 2005, Boca Raton: CRC Press. 363.
14. Mao, J.J., Translational approaches in tissue engineering and regenerative
medicine. 2008, Boston, Mass.: Artech House.
15. Nabil Dib, D.A.T., Edward B. Diethrich, editors, Stem cell therapy and tissue
engineering for cardiovascular repair [electronic resource] : from basic research
to clinical applications 2006, New York, NY: Springer.
16. Al-Rubeai, J.C.a.M., Bioreactors for tissue engineeering [electronic resource] :
principles, design and operation. Vol. 1. 2005, London: Springer.
17. Caplan, V.M.G.a.A.I., Orthopedic tissue engineering : basic science and practice
Vol. XV. 2004, New York: Marcel Dekker. 338.
18. Hatton, A.P.H.a.P.V., Biopolymer methods in tissue engineering Methods in
molecular biology. Vol. XIII. c2004, Totowa, N.J: Humana Press.
127
19. Stocum, D.L., Regenerative biology and medicine. 2006, Amsterdam ; Boston:
Elsevier Academic Press.
20. Health, N.I.o. Stem Cell Basics. Stem Cell Information 2008 [cited 2008 October
12].
21. Darwin J. Prockop, D.G.P., Bruce A. Bunnell, Mesenchymal stem cells : methods
and protocols Methods in molecular biology. 2008, Totowa, N.J.: Humana
Press,. 192.
22. Davies, J.A., Branching morphogenesis [electronic resource] Molecular biology
intelligence unit. 2006, Georgetown, Tex.: Landes Bioscience.
23. Lian, J.B. and G.S. Stein, Development of the osteoblast phenotype: molecular
mechanisms mediating osteoblast growth and differentiation. Iowa Orthop J,
1995. 15: p. 118-40.
24. Bunting, K.D., Hematopoietic stem cell protocols 2ed. Methods in molecular
biology. 2008, Totowa, N.J.: Humana
25. Growth Factor and Signal Transduction Conference (13th : 2004 : Ames, I. Stem
cell biology : development and plasticity in Growth Factor and Signal
Transduction Conference. 2004. Ames, Iowa: New York Academy of Sciences.
26. Tang, H. (1995) Mathematics of the Finite Element Method. Volume,
27. Strang, G., An Analysis of the Finite Element Method. 1973: Prentice Hall, Inc.
28. Mathematics, S.f.I.a.A. (2005) Obituary: Germund Dahlquist. Volume,
29. AB., C. (2008) COMSOL Multiphysics. Volume,
30. Teresi, A.D.S.N.S.Q.L., Modeling Bone Remodeling. Proceedings of the
COMSOL Multiphysics User's Conference 2005 Paris, 2005.
31. de Mel, A., et al., Biofunctionalization of Biomaterials for Accelerated in Situ
Endothelialization: A Review. Biomacromolecules, 2008.
32. McMahon, L.A., F.J. O'Brien, and P.J. Prendergast, Biomechanics and
mechanobiology in osteochondral tissues. Regen Med, 2008. 3(5): p. 743-59.
33. Kretlow, J.D. and A.G. Mikos, Review: mineralization of synthetic polymer
scaffolds for bone tissue engineering. Tissue Eng, 2007. 13(5): p. 927-38.
34. Abousleiman, R.I. and V.I. Sikavitsas, Bioreactors for tissues of the
musculoskeletal system. Adv Exp Med Biol, 2006. 585: p. 243-59.
35. Chromiak, J.A., et al., Bioreactor perfusion system for the long-term maintenance
of tissue-engineered skeletal muscle organoids. In Vitro Cell Dev Biol Anim,
1998. 34(9): p. 694-703.
36. Freed, L.E., et al., Tissue engineering of cartilage in space. Proc Natl Acad Sci U
S A, 1997. 94(25): p. 13885-90.
37. Freed, L.E., et al., Microgravity cultivation of cells and tissues. Gravit Space Biol
Bull, 1999. 12(2): p. 57-66.
38. G. Vunjak-Novakovic, e.a., Microgravity studies of cells and tissues,. Ann N Y
Acad Sci,, 2002. 974: p. 504-17.
39. Riddle, R.C., A.F. Taylor, and H.J. Donahue, Fluid flow assays. Methods Mol
Biol, 2008. 455: p. 335-45.
40. Alford, A.I., C.R. Jacobs, and H.J. Donahue, Oscillating fluid flow regulates gap
junction communication in osteocytic MLO-Y4 cells by an ERK1/2 MAP kinase-
dependent mechanism small star, filled. Bone, 2003. 33(1): p. 64-70.
128
41. Kim, E.J., et al., Modulating human connective tissue progenitor cell behavior on
cellulose acetate scaffolds by surface microtextures. J Biomed Mater Res A,
2008.
42. Li, X., et al., Fabrication and evaluation of calcium phosphate cement scaffold
with controlled internal channel architecture and complex shape. Proc Inst Mech
Eng [H], 2007. 221(8): p. 951-8.
43. Panda, P., et al., Stop-flow lithography to generate cell-laden microgel particles.
Lab Chip, 2008. 8(7): p. 1056-61.
44. Subramani, K. and M.A. Birch, Fabrication of poly(ethylene glycol) hydrogel
micropatterns with osteoinductive growth factors and evaluation of the effects on
osteoblast activity and function. Biomed Mater, 2006. 1(3): p. 144-54.
45. Lim, J.Y., et al., Human foetal osteoblastic cell response to polymer-demixed
nanotopographic interfaces. J R Soc Interface, 2005. 2(2): p. 97-108.
46. Caterina Minelli, I.G., Rolf Eckert, Horst Vogel, Harry Heinzelmann and Martha
Liley, Organization of nanoscale objects via polymer demixing. Colloid and
Polymer Science, 2004. 282(11): p. 1274-1278.
47. Lim, J.Y., et al., Integrin expression and osteopontin regulation in human fetal
osteoblastic cells mediated by substratum surface characteristics. Tissue Eng,
2005. 11(1-2): p. 19-29.
48. Lim, J.Y. and H.J. Donahue, Biomaterial characteristics important to skeletal
tissue engineering. J Musculoskelet Neuronal Interact, 2004. 4(4): p. 396-8.
49. Lim, J.Y. and H.J. Donahue, Cell sensing and response to micro- and
nanostructured surfaces produced by chemical and topographic patterning.
Tissue Eng, 2007. 13(8): p. 1879-91.
50. Lim, J.Y., et al., The regulation of integrin-mediated osteoblast focal adhesion
and focal adhesion kinase expression by nanoscale topography. Biomaterials,
2007. 28(10): p. 1787-97.
51. Lim, J.Y., et al., Systematic variation in osteoblast adhesion and phenotype with
substratum surface characteristics. J Biomed Mater Res A, 2004. 68(3): p. 504-
12.
52. Lim, J.Y., et al., Surface energy effects on osteoblast spatial growth and
mineralization. Biomaterials, 2008. 29(12): p. 1776-84.
53. Riddle, R.C. and H.J. Donahue, From streaming-potentials to shear stress: 25
years of bone cell mechanotransduction. J Orthop Res, 2008.
54. Riddle, R.C., K.R. Hippe, and H.J. Donahue, Chemotransport contributes to the
effect of oscillatory fluid flow on human bone marrow stromal cell proliferation. J
Orthop Res, 2008. 26(7): p. 918-24.
55. Riddle, R.C., et al., MAP kinase and calcium signaling mediate fluid flow-induced
human mesenchymal stem cell proliferation. Am J Physiol Cell Physiol, 2006.
290(3): p. C776-84.
56. Riddle, R.C., et al., ATP release mediates fluid flow-induced proliferation of
human bone marrow stromal cells. J Bone Miner Res, 2007. 22(4): p. 589-600.
57. Hansen, J.C., et al., Effect of surface nanoscale topography on elastic modulus of
individual osteoblastic cells as determined by atomic force microscopy. J
Biomech, 2007. 40(13): p. 2865-71.
129
58. Hosseinkhani, H., et al., Bone regeneration through controlled release of bone
morphogenetic protein-2 from 3-D tissue engineered nano-scaffold. J Control
Release, 2007. 117(3): p. 380-6.
59. Hosseinkhani, H., et al., Enhanced angiogenesis through controlled release of
basic fibroblast growth factor from peptide amphiphile for tissue regeneration.
Biomaterials, 2006. 27(34): p. 5836-44.
60. Hosseinkhani, H., M. Hosseinkhani, and H. Kobayashi, Proliferation and
differentiation of mesenchymal stem cells using self-assembled peptide
amphiphile nanofibers. Biomed Mater, 2006. 1(1): p. 8-15.
61. Hosseinkhani, H., et al., Osteogenic differentiation of mesenchymal stem cells in
self-assembled peptide-amphiphile nanofibers. Biomaterials, 2006. 27(22): p.
4079-86.
62. Hosseinkhani, H., et al., Ectopic bone formation in collagen sponge self-
assembled peptide-amphiphile nanofibers hybrid scaffold in a perfusion culture
bioreactor. Biomaterials, 2006. 27(29): p. 5089-98.
63. Hosseinkhani, H., et al., Bone regeneration on a collagen sponge self-assembled
peptide-amphiphile nanofiber hybrid scaffold. Tissue Eng, 2007. 13(1): p. 11-9.
64. Tian, F., et al., Quantitative analysis of cell adhesion on aligned micro- and
nanofibers. J Biomed Mater Res A, 2008. 84(2): p. 291-9.
65. Hee, C.K., M.A. Jonikas, and S.B. Nicoll, Influence of three-dimensional scaffold
on the expression of osteogenic differentiation markers by human dermal
fibroblasts. Biomaterials, 2006. 27(6): p. 875-84.
66. Holtorf, H.L., et al., Scaffold mesh size affects the osteoblastic differentiation of
seeded marrow stromal cells cultured in a flow perfusion bioreactor. J Biomed
Mater Res A, 2005. 74(2): p. 171-80.
67. Wang, T.Y. and J.H. Wu, A continuous perfusion bioreactor for long-term bone
marrow culture. Ann N Y Acad Sci, 1992. 665: p. 274-84.
68. Alvarez-Barreto, J.F., et al., Flow perfusion improves seeding of tissue
engineering scaffolds with different architectures. Ann Biomed Eng, 2007. 35(3):
p. 429-42.
69. Datta, N., et al., In vitro generated extracellular matrix and fluid shear stress
synergistically enhance 3D osteoblastic differentiation. Proc Natl Acad Sci U S
A, 2006. 103(8): p. 2488-93.
70. Gomes, M.E., et al., Influence of the porosity of starch-based fiber mesh scaffolds
on the proliferation and osteogenic differentiation of bone marrow stromal cells
cultured in a flow perfusion bioreactor. Tissue Eng, 2006. 12(4): p. 801-9.
71. Holtorf, H.L., J.A. Jansen, and A.G. Mikos, Flow perfusion culture induces the
osteoblastic differentiation of marrow stroma cell-scaffold constructs in the
absence of dexamethasone. J Biomed Mater Res A, 2005. 72(3): p. 326-34.
72. Holtorf, H.L., J.A. Jansen, and A.G. Mikos, Modulation of cell differentiation in
bone tissue engineering constructs cultured in a bioreactor. Adv Exp Med Biol,
2006. 585: p. 225-41.
73. Holtorf, H.L., et al., Flow perfusion culture of marrow stromal cells seeded on
porous biphasic calcium phosphate ceramics. Ann Biomed Eng, 2005. 33(9): p.
1238-48.
130
74. Li, X., et al., [Development of rotating perfusion bioreactor system and
application for bone tissue engineering]. Sheng Wu Yi Xue Gong Cheng Xue Za
Zhi, 2007. 24(1): p. 66-70.
75. Sikavitsas, V.I., et al., Mineralized matrix deposition by marrow stromal
osteoblasts in 3D perfusion culture increases with increasing fluid shear forces.
Proc Natl Acad Sci U S A, 2003. 100(25): p. 14683-8.
76. Sikavitsas, V.I., et al., Flow perfusion enhances the calcified matrix deposition of
marrow stromal cells in biodegradable nonwoven fiber mesh scaffolds. Ann
Biomed Eng, 2005. 33(1): p. 63-70.
77. Wang, L., et al., Flow perfusion culture of human fetal bone cells in large beta-
tricalcium phosphate scaffold with controlled architecture. J Biomed Mater Res
A, 2008.
78. Wang, Y., et al., Application of perfusion culture system improves in vitro and in
vivo osteogenesis of bone marrow-derived osteoblastic cells in porous ceramic
materials. Tissue Eng, 2003. 9(6): p. 1205-14.
79. Porter, B., et al., 3-D computational modeling of media flow through scaffolds in
a perfusion bioreactor. J Biomech, 2005. 38(3): p. 543-9.
80. Gang, E.J., et al., SSEA-4 identifies mesenchymal stem cells from bone marrow.
Blood, 2007. 109(4): p. 1743-51.
81. Ouyang, A. and S.T. Yang, A two-stage perfusion fibrous bed bioreactor system
for mass production of embryonic stem cells. Expert Opin Biol Ther, 2008. 8(7):
p. 895-909.
82. Notes, C. Bone Structure. Volume,
83. Vertebrate Embryology, Brown University.
131
APPENDIX A
ACKNOWLEDGMENTS
• Dr. Henry Donahue - I began working with Dr. Donahue in 2007, and I am very grateful
for all he has provided me as I completed this thesis. Previous work in his lab led to the
insights in oscillating fluid flow that became the cornerstone of this thesis. Without Dr.
Donahue, this thesis would not be possible.
• Dr. Peter Butler – My work with Dr. Butler spanned two summers of the Biomaterials
and Bionanotechnology Summer Institute (BBSI) and two years in the Department of
Bioengineering. He has provided me with training in COMSOL for the finite element
analyses in this thesis, and I cannot thank him enough for all he has done.
• Dr. Jung Yul Lim – I worked with Dr. Lim under the guidance of Dr. Donahue to develop
the nanoscale substrates in this thesis. Training me at first with a high level of guidance
and later without, he was instrumental in the development of my work in cell culture and
polymer science.
• Dianne McDonald – Mrs. McDonald was the acting administrator of the BBSI program,
and I worked very closely with her to develop myself during those two summers.
• Drs. Yue Zhang and Christopher Niyibizi, Jacqueline Yanoso – These two brilliant
investigators and graduate student assisted me in the flow cytometry studies of stem cells
on 2D substrates.
• Dr. Ryan Riddle, Amanda Taylor, Peter Govey – These students, among others in the
Department of Musculoskeletal Sciences, assisted me heavily in my studies.
• Dr. William Hancock – Dr. Hancock provided insights into writing techniques and
presentations necessary for the completion and defense of this thesis.
• Dr. Margaret Slattery – Dr. Slattery was instrumental in guiding me during my education
in the Department of Bioengineering.
• Carol Boring – Without Carol, I would never have graduated in time, and this thesis
would not be complete.
• Bioengineering Faculty – All of the bioengineering faculty provided me with the skills
necessary to complete this work. Dr. Keefe Manning provided a strong foundation in
biofluid mechanics, Dr. Andrew Webb provided a background in statistics, Dr. Ryan
Clement granted me training in MATLAB, and all of the other faculty/staff should be
acknowledged equally.
• SURIP, Step-Up, BBSI Students – These students, some of whom will one day be
colleagues, motivated me and kept me sane when experiments failed. They deserve more
credit than these lines deserve.
• Students of Bioengineering – My fellow students in Bioengineering at Penn State assisted
me in statistical analyses and revising this work. They were also integral in providing
positive feedback throughout the years.
• My Family – This is self-explanatory, but my family provided the most support, a level
of support that I find to be a diamond in the rough.
• All other acquaintances – Those that I met along the way, Students with whom I
interacted, Friends I have made – All of them deserve my eternal thanks for both this
thesis and a wonderful college experience.
132
APPENDIX B
All materials and information contained in this thesis are believed to be accurate and
reliable; however, the Schreyer Honors College, the Department of Bioengineering, and
The Pennsylvania State University assume no responsibility for the use of the materials
and information.
Unless otherwise noted on an individual document, the Schreyer Honors College grants
permission to copy and distribute files, documents, and information for non-commercial
use provided they are copied and distributed without alteration.
All human subjects and cell lines have been approved by the IRB at Penn State under
compliance with HIPAA.
Please note: The Final HIPAA Privacy Rule was published in August 14, 2002. All
materials dated prior to August 14, 2002 should be reviewed in conjunction with the
Final Privacy Rule published in the federal register on August 14, 2002.
133
APPENDIX C
FUNDING SOURCES
These studies have been funded jointly by the National Science Foundation (NSF) and
the National Institutes of Health (NIH). The grants established the BBSI, and research
funds from the BBSI grants were used in the completion of these studies
• NSF: EEC-0234026
• NIH: AG13087
Additionally, funding has been provided from the following sources for travel and
research:
• Department of Bioengineering
• Schreyer Honors College
• The Pennsylvania State University
• Pennsylvania Tobacco Fund
134
APPENDIX D
SUPPLEMENTAL SKETCHES
135
Figure A2. Pathways for stem cell harvesting and differentiation [83].
136
APPENDIX E
137
APPENDIX F
1. Table of Contents
• Title - COMSOL Model Report
• Table of Contents
• Model Properties
• Constants
• Geometry
• Geom1
• Solver Settings
• Postprocessing
• Variables
2. Model Properties
Property Value
Model name
Author
Company
Department
Reference
URL
Saved date Jun 19, 2008 2:00:43 PM
Creation date Jun 12, 2008 10:36:24 AM
COMSOL version COMSOL 3.4.0.248
138
Application modes and modules used in this model:
• Geom1 (2D)
o Incompressible Navier-Stokes
o Plane Stress (Structural Mechanics Module)
3. Constants
Name Expression Value Description
w 2*pi[rad/s] frequency
Patm 0[Pa]
k 40[Pa/m]
4. Geometry
Number of geometries: 1
4.1. Geom1
139
4.1.1. Point mode
140
4.1.2. Boundary mode
141
4.1.3. Subdomain mode
5. Geom1
Space dimensions: 2D
Independent variables: x, y, z
5.1. Mesh
5.1.1. Mesh Statistics
142
5.2. Application Mode: Incompressible Navier-
Stokes (ns)
Application mode type: Incompressible Navier-Stokes
Property Value
Default element type Lagrange - P2 P1
Analysis type Transient
Corner smoothing Off
Frame Frame (ref)
Weak constraints Off
Constraint type Ideal
143
5.2.2. Variables
Boundary 1 2, 4, 6-7 3
Type Inlet Wall Wall
intype p uv uv
walltype noslip noslip slip
Pressure (p0) Pa 0.5*k*cos(w*t)+Patm 0 0
Boundary 5
Type Outlet
intype uv
walltype noslip
Pressure (p0) Pa -0.5*k*cos(w*t)+Patm
Subdomain 1
Integration order (gporder) 442
Constraint order (cporder) 221
144
5.3.2. Application Mode Properties
Property Value
Default element type Lagrange - Quadratic
Analysis type Static
Large deformation On
Specify eigenvalues using Eigenfrequency
Create frame On
Deform frame Frame (deform)
Frame Frame (ref)
Weak constraints Off
Constraint type Ideal
5.3.3. Variables
6. Solver Settings
Solve using a script: off
145
6.1. Direct (PARDISO)
Solver type: Linear system solver
Parameter Value
Preordering algorithm Nested dissection
Row preordering On
Pivoting perturbation 1.0E-8
Relative tolerance 1.0E-6
Factor in error estimate 400.0
Check tolerances On
6.3. Advanced
Parameter Value
Constraint handling method Elimination
Null-space function Automatic
Assembly block size 1000
Use Hermitian transpose of constraint matrix and in symmetry Off
detection
Use complex functions with real input Off
Stop if error due to undefined operation On
Store solution on file Off
146
Type of scaling Automatic
Manual scaling
Row equilibration On
Manual control of reassembly Off
Load constant On
Constraint constant On
Mass constant On
Damping (mass) constant On
Jacobian constant On
Constraint Jacobian constant On
7. Postprocessing
147
8. Variables
8.1. Point
Name Description Unit Expression
Fxg_smps Point load in global x dir. N 0
Fyg_smps Point load in global y dir. N 0
disp_smps Total displacement m sqrt(real(u2)^2+real(v2)^2)
8.2. Boundary
8.2.1. Boundary 1-5
148
8.2.2. Boundary 6-7
8.3. Subdomain
Name Description Unit Expression
U_ns Velocity field m/s sqrt(u^2+v^2)
V_ns Vorticity 1/s vx-uy
divU_ns Divergence of 1/s ux+vy
velocity field
cellRe_ns Cell Reynolds 1 rho_ns * U_ns * h/eta_ns
number
res_u_ns Equation N/m^3 rho_ns * (ut+u * ux+v * uy)+px-F_x_ns-
residual for u eta_ns * (2 * uxx+uyy+vxy)
res_sc_u_ns Shock N/m^3 rho_ns * (ut+u * ux+v * uy)+px-F_x_ns
capturing
149
residual for u
res_v_ns Equation N/m^3 rho_ns * (vt+u * vx+v * vy)+py-F_y_ns-
residual for v eta_ns * (vxx+uyx+2 * vyy)
res_sc_v_ns Shock N/m^3 rho_ns * (vt+u * vx+v * vy)+py-F_y_ns
capturing
residual for v
beta_x_ns Convective kg/(m^2*s) rho_ns * u
field, x
component
beta_y_ns Convective kg/(m^2*s) rho_ns * v
field, y
component
Dm_ns Mean Pa*s eta_ns
diffusion
coefficient
da_ns Total time kg/m^3 rho_ns
scale factor
taum_ns GLS time- m^3*s/kg nojac(min(timestep/rho_ns,0.5 *
scale h/max(rho_ns * U_ns,6 * eta_ns/h)))
tauc_ns GLS time- m^2/s nojac(0.5 * U_ns * h * min(1,rho_ns *
scale U_ns * h/eta_ns))
Fxg_smps Body load in N/m^2 0
global x-dir.
Fyg_smps Body load in N/m^2 0
global y-dir.
disp_smps Total m sqrt(real(u2)^2+real(v2)^2)
displacement
sx_smps sx normal Pa (F11_smps * (Sx_smps *
stress global F11_smps+Sxy_smps *
sys. F12_smps)+F12_smps * (Sxy_smps *
F11_smps+Sy_smps *
F12_smps))/J_smps
sy_smps sy normal Pa (F21_smps * (Sx_smps *
stress global F21_smps+Sxy_smps *
sys. F22_smps)+F22_smps * (Sxy_smps *
F21_smps+Sy_smps *
F22_smps))/J_smps
sxy_smps sxy shear Pa (F11_smps * (Sx_smps *
stress global F21_smps+Sxy_smps *
sys. F22_smps)+F12_smps * (Sxy_smps *
F21_smps+Sy_smps *
F22_smps))/J_smps
ex_smps ex normal 1 u2x+0.5 * (u2x^2+v2x^2)
strain global
150
sys.
ey_smps ey normal 1 v2y+0.5 * (u2y^2+v2y^2)
strain global
sys.
ez_smps ez normal 1 -nu_smps * (ex_smps/((1+nu_smps) *
strain (1-2 *
nu_smps))+ey_smps/((1+nu_smps) * (1-
2 * nu_smps))) * (1+nu_smps) * (1-2 *
nu_smps)/(1-nu_smps)
exy_smps exy shear 1 0.5 * (u2y+v2x+u2x * u2y+v2x * v2y)
strain global
sys.
Sx_smps Sx Second Pa E_smps * ((1-nu_smps) *
Piola- ex_smps/((1+nu_smps) * (1-2 *
Kirchhoff nu_smps))+nu_smps *
global sys. ey_smps/((1+nu_smps) * (1-2 *
nu_smps))+nu_smps *
ez_smps/((1+nu_smps) * (1-2 *
nu_smps)))
Sy_smps Sy Second Pa E_smps * (nu_smps *
Piola- ex_smps/((1+nu_smps) * (1-2 *
Kirchhoff nu_smps))+(1-nu_smps) *
global sys. ey_smps/((1+nu_smps) * (1-2 *
nu_smps))+nu_smps *
ez_smps/((1+nu_smps) * (1-2 *
nu_smps)))
Sz_smps Sz Second Pa 0
Piola-
Kirchhoff
global sys.
Sxy_smps Sxy Second Pa E_smps * exy_smps/(1+nu_smps)
Piola-
Kirchhoff
global sys.
wz_smps Out of plane 1 if(1+2 * ez_smps<0,-1,-1+sqrt(1+2 *
derivative of ez_smps))
out-of-plane
displacement
K_smps Bulk modulus Pa E_smps/(3 * (1-2 * nu_smps))
G_smps Shear Pa 0.5 * E_smps/(1+nu_smps)
modulus
mises_smps von Mises Pa sqrt(sx_smps^2+sy_smps^2-sx_smps *
stress sy_smps+3 * sxy_smps^2)
Ws_smps Strain energy J/m^2 0.5 * thickness_smps * (ex_smps *
151
density sx_smps+ey_smps * sy_smps+2 *
exy_smps * sxy_smps)
evol_smps Volumetric strain 1 -1+Jel_smps
F11_smps Deformation 1 1+u2x
gradient 11
comp.
F12_smps Deformation 1 u2y
gradient 12
comp.
F21_smps Deformation 1 v2x
gradient 21
comp.
F22_smps Deformation 1 1+v2y
gradient 22
comp.
F33_smps Deformation 1 1+wz_smps
gradient 33
comp.
detF_smps Determinant of 1 F33_smps * (F11_smps * F22_smps-F12_smps
deformation * F21_smps)
gradient
J_smps Volume ratio 1 detF_smps
Jel_smps Elastic volume 1 J_smps
ratio
invF11_smps Inverse of 1 F22_smps * F33_smps/detF_smps
deformation
gradient 11
comp.
invF12_smps Inverse of 1 -F12_smps * F33_smps/detF_smps
deformation
gradient 12
comp.
invF21_smps Inverse of 1 -F21_smps * F33_smps/detF_smps
deformation
gradient 21
comp.
invF22_smps Inverse of 1 F11_smps * F33_smps/detF_smps
deformation
gradient 22
comp.
invF33_smps Inverse of 1 (F11_smps * F22_smps-F21_smps *
deformation F12_smps)/detF_smps
gradient 33
comp.
sz_smps sz normal stress Pa 0
global sys.
tresca_smps Tresca stress Pa max(max(abs(s1_smps-
s2_smps),abs(s2_smps-
s3_smps)),abs(s1_smps-s3_smps))
152
COMSOL Model Report – Cell Confluence Study
1. Table of Contents
• Title - COMSOL Model Report
• Table of Contents
• Model Properties
• Geometry
• Geom1
• Solver Settings
• Postprocessing
• Variables
2. Model Properties
Property Value
Model name
Author
Company
Department
Reference
URL
Saved date Jun 6, 2008 2:06:39 PM
Creation date Jun 6, 2008 1:31:37 PM
COMSOL version COMSOL 3.4.0.248
• Geom1 (2D)
o Plane Strain (Structural Mechanics Module)
o Incompressible Navier-Stokes
153
3. Geometry
Number of geometries: 1
3.1. Geom1
154
3.1.1. Point mode
155
3.1.2. Boundary mode
156
3.1.3. Subdomain mode
4. Geom1
Space dimensions: 2D
Independent variables: x, y, z
4.1. Mesh
4.1.1. Mesh Statistics
157
4.2. Application Mode: Plane Strain (smpn)
Application mode type: Plane Strain (Structural Mechanics Module)
Property Value
Default element type Lagrange - Quadratic
Analysis type Static
Large deformation On
Specify eigenvalues using Eigenfrequency
Create frame Off
158
Deform frame Frame (ref)
Frame Frame (ref)
Weak constraints Off
Constraint type Ideal
4.2.3. Variables
Subdomain 1
name Solid domain
Property Value
Default element type Lagrange - P2 P1
Analysis type Stationary
Corner smoothing Off
Frame Frame (ref)
Weak constraints Off
Constraint type Ideal
4.3.2. Variables
159
4.3.3. Boundary Settings
Subdomain 1
Integration order (gporder) 442
Constraint order (cporder) 221
5. Solver Settings
Solve using a script: off
Parameter Value
Preordering algorithm Nested dissection
Row preordering On
Pivoting perturbation 1.0E-8
Relative tolerance 1.0E-6
Factor in error estimate 400.0
Check tolerances On
160
5.2. Stationary
Parameter Value
Linearity Automatic
Relative tolerance 1.0E-6
Maximum number of iterations 25
Manual tuning of damping parameters Off
Highly nonlinear problem On
Initial damping factor 1.0
Minimum damping factor 1.0E-4
Restriction for step size update 10.0
5.3. Advanced
Parameter Value
Constraint handling method Elimination
Null-space function Automatic
Assembly block size 1000
Use Hermitian transpose of constraint matrix and in symmetry Off
detection
Use complex functions with real input Off
Stop if error due to undefined operation On
Store solution on file Off
Type of scaling None
Manual scaling
Row equilibration On
Manual control of reassembly Off
Load constant On
Constraint constant On
Mass constant On
Damping (mass) constant On
Jacobian constant On
Constraint Jacobian constant On
161
6. Postprocessing
7. Variables
7.1. Point
Name Description Unit Expression
Fxg_smpn Point load in global x dir. N 0
Fyg_smpn Point load in global y dir. N 0
disp_smpn Total displacement m sqrt(real(u2)^2+real(v2)^2)
7.2. Boundary
Name Description Unit Expression
Fxg_smpn Edge load in N/m 0
global x-dir.
Fyg_smpn Edge load in N/m 0
global y-dir.
162
disp_smpn Total m sqrt(real(u2)^2+real(v2)^2)
displacement
Tax_smpn Surface traction Pa (F11_smpn * Sx_smpn+F12_smpn *
(force/area) in x Sxy_smpn) * nx_smpn+(F11_smpn *
dir. Sxy_smpn+F12_smpn * Sy_smpn) *
ny_smpn
Tay_smpn Surface traction Pa (F21_smpn * Sx_smpn+F22_smpn *
(force/area) in y Sxy_smpn) * nx_smpn+(F21_smpn *
dir. Sxy_smpn+F22_smpn * Sy_smpn) *
ny_smpn
K_x_ns Viscous force per Pa eta_ns * (2 * nx_ns * ux+ny_ns * (uy+vx))
area, x
component
T_x_ns Total force per Pa -nx_ns * p+2 * nx_ns * eta_ns * ux+ny_ns *
area, x eta_ns * (uy+vx)
component
K_y_ns Viscous force per Pa eta_ns * (nx_ns * (vx+uy)+2 * ny_ns * vy)
area, y
component
T_y_ns Total force per Pa -ny_ns * p+nx_ns * eta_ns * (vx+uy)+2 *
area, y ny_ns * eta_ns * vy
component
7.3. Subdomain
Name Description Unit Expression
Fxg_smpn Body load in N/m^2 0
global x-dir.
Fyg_smpn Body load in N/m^2 0
global y-dir.
disp_smpn Total m sqrt(real(u2)^2+real(v2)^2)
displacemen
t
sx_smpn sx normal Pa (F11_smpn * (Sx_smpn *
stress global F11_smpn+Sxy_smpn *
sys. F12_smpn)+F12_smpn * (Sxy_smpn *
F11_smpn+Sy_smpn *
F12_smpn))/J_smpn
sy_smpn sy normal Pa (F21_smpn * (Sx_smpn *
stress global F21_smpn+Sxy_smpn *
sys. F22_smpn)+F22_smpn * (Sxy_smpn *
F21_smpn+Sy_smpn *
F22_smpn))/J_smpn
sz_smpn sz normal Pa Sz_smpn * F33_smpn^2/J_smpn
163
stress global
sys.
sxy_smpn sxy shear Pa (F11_smpn * (Sx_smpn *
stress global F21_smpn+Sxy_smpn *
sys. F22_smpn)+F12_smpn * (Sxy_smpn *
F21_smpn+Sy_smpn *
F22_smpn))/J_smpn
ex_smpn ex normal 1 u2x+0.5 * (u2x^2+v2x^2)
strain global
sys.
ey_smpn ey normal 1 v2y+0.5 * (u2y^2+v2y^2)
strain global
sys.
exy_smpn exy shear 1 0.5 * (u2y+v2x+u2x * u2y+v2x * v2y)
strain global
sys.
Sx_smpn Sx Second Pa E_smpn * ((1-nu_smpn) *
Piola- ex_smpn/((1+nu_smpn) * (1-2 *
Kirchhoff nu_smpn))+nu_smpn *
global sys. ey_smpn/((1+nu_smpn) * (1-2 *
nu_smpn)))
Sy_smpn Sy Second Pa E_smpn * (nu_smpn *
Piola- ex_smpn/((1+nu_smpn) * (1-2 *
Kirchhoff nu_smpn))+(1-nu_smpn) *
global sys. ey_smpn/((1+nu_smpn) * (1-2 *
nu_smpn)))
Sz_smpn Sz Second Pa E_smpn * nu_smpn *
Piola- (ex_smpn/((1+nu_smpn) * (1-2 *
Kirchhoff nu_smpn))+ey_smpn/((1+nu_smpn) * (1-
global sys. 2 * nu_smpn)))
Sxy_smpn Sxy Second Pa E_smpn * exy_smpn/(1+nu_smpn)
Piola-
Kirchhoff
global sys.
K_smpn Bulk Pa E_smpn/(3 * (1-2 * nu_smpn))
modulus
G_smpn Shear Pa 0.5 * E_smpn/(1+nu_smpn)
modulus
mises_smpn von Mises Pa sqrt(sx_smpn^2+sy_smpn^2+sz_smpn^2
stress -sx_smpn * sy_smpn-sy_smpn *
sz_smpn-sx_smpn * sz_smpn+3 *
sxy_smpn^2)
Ws_smpn Strain J/m^2 0.5 * thickness_smpn * (ex_smpn *
energy sx_smpn+ey_smpn * sy_smpn+2 *
164
density exy_smpn * sxy_smpn)
evol_smpn Volumetric 1 -1+Jel_smpn
strain
ez_smpn ez normal 1 0
strain global
sys.
F11_smpn Deformation 1 1+u2x
gradient 11
comp.
F12_smpn Deformation 1 u2y
gradient 12
comp.
F21_smpn Deformation 1 v2x
gradient 21
comp.
F22_smpn Deformation 1 1+v2y
gradient 22
comp.
F33_smpn Deformation 1 1+ez_smpn
gradient 33
comp.
detF_smpn Determinant 1 F33_smpn * (F11_smpn * F22_smpn-
of F12_smpn * F21_smpn)
deformation
gradient
J_smpn Volume ratio 1 detF_smpn
Jel_smpn Elastic 1 J_smpn
volume ratio
invF11_smp Inverse of 1 F22_smpn * F33_smpn/detF_smpn
n deformation
gradient 11
comp.
invF12_smp Inverse of 1 -F12_smpn * F33_smpn/detF_smpn
n deformation
gradient 12
comp.
invF21_smp Inverse of 1 -F21_smpn * F33_smpn/detF_smpn
n deformation
gradient 21
comp.
invF22_smp Inverse of 1 F11_smpn * F33_smpn/detF_smpn
n deformation
gradient 22
comp.
165
invF33_smp Inverse of 1 (F11_smpn * F22_smpn-F21_smpn *
n deformation F12_smpn)/detF_smpn
gradient 33
comp.
tresca_smpn Tresca Pa max(max(abs(s1_smpn-
stress s2_smpn),abs(s2_smpn-
s3_smpn)),abs(s1_smpn-s3_smpn))
U_ns Velocity field m/s sqrt(u^2+v^2)
V_ns Vorticity 1/s vx-uy
divU_ns Divergence 1/s ux+vy
of velocity
field
cellRe_ns Cell 1 rho_ns * U_ns * h/eta_ns
Reynolds
number
res_u_ns Equation N/m^3 rho_ns * (u * ux+v * uy)+px-F_x_ns-
residual for u eta_ns * (2 * uxx+uyy+vxy)
res_sc_u_ns Shock N/m^3 rho_ns * (u * ux+v * uy)+px-F_x_ns
capturing
residual for u
res_v_ns Equation N/m^3 rho_ns * (u * vx+v * vy)+py-F_y_ns-
residual for v eta_ns * (vxx+uyx+2 * vyy)
res_sc_v_ns Shock N/m^3 rho_ns * (u * vx+v * vy)+py-F_y_ns
capturing
residual for v
beta_x_ns Convective kg/(m^2*s rho_ns * u
field, x )
component
beta_y_ns Convective kg/(m^2*s rho_ns * v
field, y )
component
Dm_ns Mean Pa*s eta_ns
diffusion
coefficient
da_ns Total time kg/m^3 rho_ns
scale factor
taum_ns GLS time- m^3*s/kg nojac(0.5 * h/max(rho_ns * U_ns,6 *
scale eta_ns/h))
tauc_ns GLS time- m^2/s nojac(0.5 * U_ns * h * min(1,rho_ns *
scale U_ns * h/eta_ns))
166
COMSOL Model Report – Young’s Modulus Study
1. Table of Contents
• Title - COMSOL Model Report
• Table of Contents
• Model Properties
• Constants
• Geometry
• Geom1
• Solver Settings
• Postprocessing
• Variables
2. Model Properties
Property Value
Model name
Author
Company
Department
Reference
URL
Saved date Jul 9, 2008 1:32:46 PM
Creation date Jun 12, 2008 10:36:24 AM
COMSOL version COMSOL 3.4.0.248
• Geom1 (2D)
o Incompressible Navier-Stokes
o Plane Stress (Structural Mechanics Module)
167
3. Constants
Name Expression Value Description
w 2*pi[rad/s] frequency
Patm 0[Pa]
k 40[Pa/m]
4. Geometry
Number of geometries: 1
4.1. Geom1
168
4.1.1. Point mode
169
4.1.2. Boundary mode
170
4.1.3. Subdomain mode
5. Geom1
Space dimensions: 2D
Independent variables: x, y, z
5.1. Mesh
5.1.1. Mesh Statistics
171
5.2. Application Mode: Incompressible Navier-
Stokes (ns)
Application mode type: Incompressible Navier-Stokes
Property Value
Default element type Lagrange - P2 P1
Analysis type Transient
Corner smoothing Off
Frame Frame (ref)
Weak constraints Off
Constraint type Ideal
172
5.2.2. Variables
Boundary 1 2, 4, 6-7 3
Type Inlet Wall Wall
intype p uv uv
walltype noslip noslip slip
Pressure (p0) Pa 0.5*k*cos(w*t)+Patm 0 0
Boundary 5
Type Outlet
intype uv
walltype noslip
Pressure (p0) Pa -0.5*k*cos(w*t)+Patm
Subdomain 1
Integration order (gporder) 442
Constraint order (cporder) 221
Property Value
173
Default element type Lagrange - Quadratic
Analysis type Static
Large deformation On
Specify eigenvalues using Eigenfrequency
Create frame On
Deform frame Frame (deform)
Frame Frame (ref)
Weak constraints Off
Constraint type Ideal
5.3.3. Variables
Subdomain 1
Young's modulus (E) Pa 9000
6. Solver Settings
Solve using a script: off
174
175
6.1. Direct (PARDISO)
Solver type: Linear system solver
Parameter Value
Preordering algorithm Nested dissection
Row preordering On
Pivoting perturbation 1.0E-8
Relative tolerance 1.0E-6
Factor in error estimate 400.0
Check tolerances On
6.3. Advanced
Parameter Value
Constraint handling method Elimination
Null-space function Automatic
Assembly block size 1000
Use Hermitian transpose of constraint matrix and in symmetry Off
detection
Use complex functions with real input Off
Stop if error due to undefined operation On
Store solution on file Off
176
Type of scaling Automatic
Manual scaling
Row equilibration On
Manual control of reassembly Off
Load constant On
Constraint constant On
Mass constant On
Damping (mass) constant On
Jacobian constant On
Constraint Jacobian constant On
7. Postprocessing
177
8. Variables
8.1. Point
Name Description Unit Expression
Fxg_smps Point load in global x dir. N 0
Fyg_smps Point load in global y dir. N 0
disp_smps Total displacement m sqrt(real(u2)^2+real(v2)^2)
8.2. Boundary
8.2.1. Boundary 1-5
178
8.2.2. Boundary 6-7
8.3. Subdomain
Name Description Unit Expression
U_ns Velocity field m/s sqrt(u^2+v^2)
V_ns Vorticity 1/s vx-uy
divU_ns Divergence of 1/s ux+vy
velocity field
cellRe_ns Cell Reynolds 1 rho_ns * U_ns * h/eta_ns
number
res_u_ns Equation N/m^3 rho_ns * (ut+u * ux+v * uy)+px-F_x_ns-
residual for u eta_ns * (2 * uxx+uyy+vxy)
res_sc_u_ns Shock N/m^3 rho_ns * (ut+u * ux+v * uy)+px-F_x_ns
capturing
179
residual for u
res_v_ns Equation N/m^3 rho_ns * (vt+u * vx+v * vy)+py-F_y_ns-
residual for v eta_ns * (vxx+uyx+2 * vyy)
res_sc_v_ns Shock N/m^3 rho_ns * (vt+u * vx+v * vy)+py-F_y_ns
capturing
residual for v
beta_x_ns Convective kg/(m^2*s) rho_ns * u
field, x
component
beta_y_ns Convective kg/(m^2*s) rho_ns * v
field, y
component
Dm_ns Mean diffusion Pa*s eta_ns
coefficient
da_ns Total time kg/m^3 rho_ns
scale factor
taum_ns GLS time- m^3*s/kg nojac(min(timestep/rho_ns,0.5 *
scale h/max(rho_ns * U_ns,6 * eta_ns/h)))
tauc_ns GLS time- m^2/s nojac(0.5 * U_ns * h * min(1,rho_ns *
scale U_ns * h/eta_ns))
Fxg_smps Body load in N/m^2 0
global x-dir.
Fyg_smps Body load in N/m^2 0
global y-dir.
disp_smps Total m sqrt(real(u2)^2+real(v2)^2)
displacement
sx_smps sx normal Pa (F11_smps * (Sx_smps *
stress global F11_smps+Sxy_smps *
sys. F12_smps)+F12_smps * (Sxy_smps *
F11_smps+Sy_smps *
F12_smps))/J_smps
sy_smps sy normal Pa (F21_smps * (Sx_smps *
stress global F21_smps+Sxy_smps *
sys. F22_smps)+F22_smps * (Sxy_smps *
F21_smps+Sy_smps *
F22_smps))/J_smps
sxy_smps sxy shear stress Pa (F11_smps * (Sx_smps *
global sys. F21_smps+Sxy_smps *
F22_smps)+F12_smps * (Sxy_smps *
F21_smps+Sy_smps *
F22_smps))/J_smps
ex_smps ex normal strain 1 u2x+0.5 * (u2x^2+v2x^2)
global sys.
ey_smps ey normal strain 1 v2y+0.5 * (u2y^2+v2y^2)
180
global sys.
ez_smps ez normal strain 1 -nu_smps * (ex_smps/((1+nu_smps) * (1-2
* nu_smps))+ey_smps/((1+nu_smps) * (1-2
* nu_smps))) * (1+nu_smps) * (1-2 *
nu_smps)/(1-nu_smps)
exy_smps exy shear strain 1 0.5 * (u2y+v2x+u2x * u2y+v2x * v2y)
global sys.
Sx_smps Sx Second Pa E_smps * ((1-nu_smps) *
Piola-Kirchhoff ex_smps/((1+nu_smps) * (1-2 *
global sys. nu_smps))+nu_smps *
ey_smps/((1+nu_smps) * (1-2 *
nu_smps))+nu_smps *
ez_smps/((1+nu_smps) * (1-2 * nu_smps)))
Sy_smps Sy Second Pa E_smps * (nu_smps *
Piola-Kirchhoff ex_smps/((1+nu_smps) * (1-2 *
global sys. nu_smps))+(1-nu_smps) *
ey_smps/((1+nu_smps) * (1-2 *
nu_smps))+nu_smps *
ez_smps/((1+nu_smps) * (1-2 * nu_smps)))
Sz_smps Sz Second Pa 0
Piola-Kirchhoff
global sys.
Sxy_smps Sxy Second Pa E_smps * exy_smps/(1+nu_smps)
Piola-Kirchhoff
global sys.
wz_smps Out of plane 1 if(1+2 * ez_smps<0,-1,-1+sqrt(1+2 *
derivative of ez_smps))
out-of-plane
displacement
K_smps Bulk modulus Pa E_smps/(3 * (1-2 * nu_smps))
G_smps Shear modulus Pa 0.5 * E_smps/(1+nu_smps)
mises_smps von Mises Pa sqrt(sx_smps^2+sy_smps^2-sx_smps *
stress sy_smps+3 * sxy_smps^2)
Ws_smps Strain energy J/m^2 0.5 * thickness_smps * (ex_smps *
density sx_smps+ey_smps * sy_smps+2 *
exy_smps * sxy_smps)
evol_smps Volumetric 1 -1+Jel_smps
strain
F11_smps Deformation 1 1+u2x
gradient 11
comp.
F12_smps Deformation 1 u2y
gradient 12
comp.
F21_smps Deformation 1 v2x
gradient 21
comp.
F22_smps Deformation 1 1+v2y
181
gradient 22
comp.
F33_smps Deformation 1 1+wz_smps
gradient 33
comp.
detF_smps Determinant of 1 F33_smps * (F11_smps * F22_smps-
deformation F12_smps * F21_smps)
gradient
J_smps Volume ratio 1 detF_smps
Jel_smps Elastic volume 1 J_smps
ratio
invF11_smps Inverse of 1 F22_smps * F33_smps/detF_smps
deformation
gradient 11
comp.
invF12_smps Inverse of 1 -F12_smps * F33_smps/detF_smps
deformation
gradient 12
comp.
invF21_smps Inverse of 1 -F21_smps * F33_smps/detF_smps
deformation
gradient 21
comp.
invF22_smps Inverse of 1 F11_smps * F33_smps/detF_smps
deformation
gradient 22
comp.
invF33_smps Inverse of 1 (F11_smps * F22_smps-F21_smps *
deformation F12_smps)/detF_smps
gradient 33
comp.
sz_smps sz normal stress Pa 0
global sys.
tresca_smps Tresca stress Pa max(max(abs(s1_smps-
s2_smps),abs(s2_smps-
s3_smps)),abs(s1_smps-s3_smps))
182
COMSOL Model Report – Bioreactor Design
1. Table of Contents
• Title - COMSOL Model Report
• Table of Contents
• Model Properties
• Constants
• Geometry
• Geom1
• Geom2
• Materials/Coefficients Library
• Solver Settings
• Postprocessing
• Variables
2. Model Properties
Property Value
Model name
Author
Company
Department
Reference
URL
Saved date Jun 24, 2008 2:22:41 PM
Creation date Jun 9, 2008 1:27:09 PM
COMSOL version COMSOL 3.4.0.248
183
Application modes and modules used in this model:
3. Constants
Name Expression Value Description
T 273[K] temperature
4. Geometry
Number of geometries: 2
4.1. Geom1
184
4.1.1. Point mode
185
4.1.2. Boundary mode
186
4.1.3. Subdomain mode
187
4.2. Geom2
188
4.2.1. Point mode
189
4.2.2. Edge mode
190
4.2.3. Boundary mode
191
4.2.4. Subdomain mode
5. Geom1
Space dimensions: Axial symmetry (2D)
5.1. Mesh
5.1.1. Mesh Statistics
192
5.2. Application Mode: Axial Symmetry, Stress-
Strain (smaxi)
Application mode type: Axial Symmetry, Stress-Strain (Structural Mechanics
Module)
193
5.2.2. Application Mode Properties
Property Value
Default element type Lagrange - Quadratic
Analysis type Static
Large deformation On
Specify eigenvalues using Eigenfrequency
Create frame Off
Deform frame Frame (ref)
Frame Frame (ref)
Weak constraints Off
Constraint type Ideal
5.2.3. Variables
194
5.3. Application Mode: Incompressible Navier-
Stokes (chns)
Application mode type: Incompressible Navier-Stokes (Chemical Engineering
Module)
Property Value
Default element type Lagrange - P2 P1
Analysis type Stationary
Corner smoothing Off
Weakly compressible flow Off
Turbulence model None
Realizability Off
Non-Newtonian flow Off
Brinkman on by default Off
Two-phase flow Single-phase flow
Swirl velocity Off
Frame Frame (ref)
Weak constraints Off
Constraint type Ideal
5.3.2. Variables
Dependent variables: u4, v4, w2, p4, logk2, logd2, logw2, phi2, nrw2, nzw2
Boundary 1, 3, 5, 7, 9, 11, 13 2 15
Type Symmetry boundary Open boundary Inlet
Normal inflow velocity (U0in) m/s 1 1 0.03
Volume per time unit (V0) m3/s 0 0 0.016
flowtype velocity velocity volume
Boundary 16-23, 25
195
Type Wall
Normal inflow velocity (U0in) m/s 1
Volume per time unit (V0) m3/s 0
flowtype velocity
6. Geom2
Space dimensions: 3D
Independent variables: x, y, z
6.1. Mesh
6.1.1. Mesh Statistics
196
7. Materials/Coefficients Library
7.1. PMMA
Parameter Value
Heat capacity at constant pressure (C) 1420[J/(kg*K)]
Young's modulus (E) 3e9[Pa]
Thermal expansion coeff. (alpha) 70e-6[1/K]
Relative permittivity (epsilonr) 3.0
Thermal conductivity (k) 0.19[W/(m*K)]
Poisson's ratio (nu) 0.40
Density (rho) 1190[kg/m^3]
7.2. Polyethylene
Parameter Value
Heat capacity at constant pressure (C) 1900[J/(kg*K)]
Young's modulus (E) 1e9[Pa]
197
Thermal expansion coeff. (alpha) 150e-6[1/K]
Relative permittivity (epsilonr) 2.3
Thermal conductivity (k) 0.38[W/(m*K)]
Density (rho) 930[kg/m^3]
7.3.1. Functions
8. Solver Settings
Solve using a script: off
198
8.1. Direct (UMFPACK)
Solver type: Linear system solver
Parameter Value
Pivot threshold 0.1
Memory allocation factor 0.7
8.2. Stationary
Parameter Value
Linearity Automatic
Relative tolerance 1.0E-6
Maximum number of iterations 25
Manual tuning of damping parameters Off
Highly nonlinear problem Off
Initial damping factor 1.0
Minimum damping factor 1.0E-4
Restriction for step size update 10.0
8.3. Advanced
Parameter Value
Constraint handling method Elimination
Null-space function Automatic
Assembly block size 1000
Use Hermitian transpose of constraint matrix and in symmetry detection Off
Use complex functions with real input Off
Stop if error due to undefined operation On
Store solution on file Off
Type of scaling None
Manual scaling
Row equilibration On
Manual control of reassembly Off
Load constant On
Constraint constant On
Mass constant On
Damping (mass) constant On
Jacobian constant On
Constraint Jacobian constant On
199
9. Postprocessing
200
10. Variables
10.1. Point
Name Description Unit Expression
FRg_smaxi Point load in global R dir. N 0
FZg_smaxi Point load in global Z dir. N 0
disp_smaxi Total displacement m sqrt(real(uaxi_smaxi)^2+real(w)^2)
uaxi_smaxi R-displacement m uor * R
uaxiR_smaxi R derivative of R- 1 uorR * R+uor
displacement
uaxiZ_smaxi Z derivative of R 1 uorZ * R
displacement
uaxi_t_smaxi R-velocity m/s diff(uaxi_smaxi,t)
201
10.2. Boundary
10.2.1. Boundary 1-9, 11-25
202
10.2.2. Boundary 10
203
10.3. Subdomain
Name Descriptio Unit Expression
n
FRg_smaxi Body load N/m^3 0
in global R-
dir.
FZg_smaxi Body load N/m^3 0
in global Z-
dir.
disp_smaxi Total m sqrt(real(uaxi_smaxi)^2+real(w)^2)
displaceme
nt
uaxi_smaxi R- m uor * R
displaceme
nt
uaxiR_smaxi R derivative 1 uorR * R+uor
of R-
displaceme
nt
uaxiZ_smaxi Z derivative 1 uorZ * R
of R
displaceme
nt
uaxi_t_smaxi R-velocity m/s diff(uaxi_smaxi,t)
sR_smaxi sR normal Pa (F11_smaxi * (SR_smaxi *
stress F11_smaxi+SRZ_smaxi *
global sys. F13_smaxi)+F13_smaxi * (SRZ_smaxi *
F11_smaxi+SZ_smaxi *
F13_smaxi))/J_smaxi
sZ_smaxi sZ normal Pa (F31_smaxi * (SR_smaxi *
stress F31_smaxi+SRZ_smaxi *
global sys. F33_smaxi)+F33_smaxi * (SRZ_smaxi *
F31_smaxi+SZ_smaxi *
F33_smaxi))/J_smaxi
sPHI_smaxi sPHI Pa SPHI_smaxi * F22_smaxi^2/J_smaxi
normal
stress
sRZ_smaxi sRZ shear Pa (F11_smaxi * (SR_smaxi *
stress F31_smaxi+SRZ_smaxi *
global sys. F33_smaxi)+F13_smaxi * (SRZ_smaxi *
F31_smaxi+SZ_smaxi *
F33_smaxi))/J_smaxi
eR_smaxi eR normal 1 uorR * R+uor+0.5 * ((uorR *
204
strain global R+uor)^2+wR^2)
sys.
eZ_smaxi eZ normal 1 wZ+0.5 * ((uorZ * R)^2+wZ^2)
strain global
sys.
ePHI_smaxi ePHI 1 uor+0.5 * uor^2
normal
strain
eRZ_smaxi eRZ shear 1 0.5 * (uorZ * R+wR+(uorR * R+uor) * uorZ
strain global * R+wR * wZ)
sys.
SR_smaxi SR Second Pa E_smaxi * ((1-nu_smaxi) *
Piola- eR_smaxi/((1+nu_smaxi) * (1-2 *
Kirchhoff nu_smaxi))+nu_smaxi *
global sys. ePHI_smaxi/((1+nu_smaxi) * (1-2 *
nu_smaxi))+nu_smaxi *
eZ_smaxi/((1+nu_smaxi) * (1-2 *
nu_smaxi)))
SZ_smaxi SZ Second Pa E_smaxi * (nu_smaxi *
Piola- eR_smaxi/((1+nu_smaxi) * (1-2 *
Kirchhoff nu_smaxi))+nu_smaxi *
global sys. ePHI_smaxi/((1+nu_smaxi) * (1-2 *
nu_smaxi))+(1-nu_smaxi) *
eZ_smaxi/((1+nu_smaxi) * (1-2 *
nu_smaxi)))
SPHI_smaxi SPHI Pa E_smaxi * (nu_smaxi *
Second eR_smaxi/((1+nu_smaxi) * (1-2 *
Piola- nu_smaxi))+(1-nu_smaxi) *
Kirchhoff ePHI_smaxi/((1+nu_smaxi) * (1-2 *
nu_smaxi))+nu_smaxi *
eZ_smaxi/((1+nu_smaxi) * (1-2 *
nu_smaxi)))
SRZ_smaxi SRZ Pa E_smaxi * eRZ_smaxi/(1+nu_smaxi)
Second
Piola-
Kirchhoff
global sys.
K_smaxi Bulk Pa E_smaxi/(3 * (1-2 * nu_smaxi))
modulus
G_smaxi Shear Pa 0.5 * E_smaxi/(1+nu_smaxi)
modulus
mises_smaxi von Mises Pa sqrt(sR_smaxi^2+sPHI_smaxi^2+sZ_smax
stress i^2-sR_smaxi * sPHI_smaxi-sPHI_smaxi *
sZ_smaxi-sR_smaxi * sZ_smaxi+3 *
sRZ_smaxi^2)
205
Ws_smaxi Strain J/m^3 0.5 * (eR_smaxi * sR_smaxi+ePHI_smaxi *
energy sPHI_smaxi+eZ_smaxi * sZ_smaxi+2 *
density eRZ_smaxi * sRZ_smaxi)
evol_smaxi Volumetric 1 -1+Jel_smaxi
strain
F11_smaxi Deformation 1 1+uaxiR_smaxi
gradient 11
comp.
F13_smaxi Deformation 1 uaxiZ_smaxi
gradient 13
comp.
F22_smaxi Deformation 1 1+uor
gradient 22
comp.
F31_smaxi Deformation 1 wR
gradient 31
comp.
F33_smaxi Deformation 1 1+wZ
gradient 33
comp.
detF_smaxi Determinant 1 F22_smaxi * (F11_smaxi * F33_smaxi-
of F13_smaxi * F31_smaxi)
deformation
gradient
J_smaxi Volume 1 detF_smaxi
ratio
Jel_smaxi Elastic 1 J_smaxi
volume ratio
invF11_smaxi Inverse of 1 F22_smaxi * F33_smaxi/detF_smaxi
deformation
gradient 11
comp.
invF13_smaxi Inverse of 1 -F13_smaxi * F22_smaxi/detF_smaxi
deformation
gradient 13
comp.
invF22_smaxi Inverse of 1 (F11_smaxi * F33_smaxi-F31_smaxi *
deformation F13_smaxi)/detF_smaxi
gradient 22
comp.
invF31_smaxi Inverse of 1 -F31_smaxi * F22_smaxi/detF_smaxi
deformation
gradient 31
comp.
206
invF33_smaxi Inverse of 1 F11_smaxi * F22_smaxi/detF_smaxi
deformation
gradient 33
comp.
tresca_smaxi Tresca Pa max(max(abs(s1_smaxi-
stress s2_smaxi),abs(s2_smaxi-
s3_smaxi)),abs(s1_smaxi-s3_smaxi))
U_chns Velocity m/s sqrt(u4^2+v4^2)
field
V_chns Vorticity 1/s u4Z-v4R
divU_chns Divergence 1/s u4R+v4Z+u4/R
of velocity
field
cellRe_chns Cell 1 rho_chns * U_chns * h/eta_chns
Reynolds
number
res_u4_chns Equation Pa R * (rho_chns * (u4 * u4R+v4 * u4Z)+p4R-
residual for F_r_chns)+2 * eta_chns * (u4/R-u4R)-
u4 eta_chns * R * (2 * u4RR+u4ZZ+v4RZ)
res_sc_u4_ch Shock Pa R * (rho_chns * (u4 * u4R+v4 * u4Z)+p4R-
ns capturing F_r_chns)+2 * eta_chns * (u4/R-u4R)
residual for
u4
res_v4_chns Equation Pa R * (rho_chns * (u4 * v4R+v4 * v4Z)+p4Z-
residual for F_z_chns)-eta_chns * (R *
v4 (v4RR+u4ZR)+2 * R * v4ZZ+u4Z+v4R)
res_sc_v4_chns Shock Pa R * (rho_chns * (u4 * v4R+v4 * v4Z)+p4Z-
capturing F_z_chns)
residual for
v4
beta_R_chns Convective Pa*s R * rho_chns * u4
field, R
component
beta_Z_chns Convective Pa*s R * rho_chns * v4
field, Z
component
Dm_chns Mean kg/s R * eta_chns
diffusion
coefficient
da_chns Total time kg/m^2 R * rho_chns
scale factor
taum_chns GLS time- m^3*s/k nojac(0.5 * h/max(rho_chns * U_chns,6 *
scale g eta_chns/h))
tauc_chns GLS time- m^2/s nojac(0.5 * U_chns * h * min(1,rho_chns *
scale U_chns * h/eta_chns))
207
APPENDIX G
Biographical Information
Education
Thesis Title: The Novel Design of a Bioreactor for in vitro Proliferation and
Differentiation of Human Mesenchymal Stem Cells
Thesis Advisors: Henry J. Donahue (Hershey, PA)
Peter J. Butler (University Park, PA)
Work Experience
208
01/2006-05/2009 Penn State Learning Center
SI Leader, Tutor, Teaching Assistant
Assisted students in CHEM 110(12), BIOE 201, PHYS 211, and
MATH 140
Supervisor: Janice Smith, Herbert Lipowsky, William Hancock
Research
Conferences: Joshua D. Salvi, Jung Yul Lim, Yue Zhang, Jacqueline Yanoso,
Christopher Niyibizi, Henry J. Donahue, Culture on specific
nanoscale topographies selects for subpopulations of stem cells
209
with increased osteogenic potential. 55th Orthopaedic Research
Society (ORS), February 22-25, 2009, Las Vegas, Nevada, USA.
Extracurricular Involvement
Related Skills
Programming: C++/C#
JavaScript, HTML
MATLAB
210