<<

The design of a novel 3D micropatterned tumor microenvironment model to study

the intercellular effects of and cancer cells

A Thesis

Submitted to the Faculty

of

Drexel University

By

Christine Maylin Ho

in partial fulfillment of the

requirements for the degree

of

Master of Science in Biomedical Engineering

August 2013

© Copyright 2013 Christine M. Ho. All Rights Reserved.

iii

ACKNOWLEDGEMENTS

I would like to thank the following people who were instrumental in the success of this thesis:

Kenneth Barbee, Ph.D and Alisa Morss Clyne, Ph.D for serving on my thesis committee,

Michael Yang, Ph.D and Anant Chopra, Ph.D, from the Microfabrication Laboratory of Christopher S. Chen, Ph.D, and Rebecca Urbano, Ph.D candidate, from the Vascular Kinetics Laboratory of Alisa Morss Clyne, Ph.D for assistance and instructions on specifics of my project,

And I am especially grateful to Adrian Shieh, Ph.D for his guidance and teachings, and the rest of the Laboratory of Tumor Mechanobiology and Microenvironment for their constant support and encouragement. In particular, thank you Alima Tchafa, Ph.D candidate, and Arpit Shah, Ph.D candidate.

iv

TABLE OF CONTENTS ACKNOWLEDGEMENTS ...... iii LIST OF FIGURES ...... vi ABSTRACT ...... viii CHAPTER 1: INTRODUCTION ...... 1 CHAPTER 2: BACKGROUND ...... 3 2.1 Tumor microenvironment ...... 3 2.2 Cancer associated fibroblasts (CAFs) ...... 4 2.3 Effects of soluble factors on the tumor environment ...... 6 2.4 Mechanical effects on the tumor environment ...... 7 2.5 Comparison of 2D and 3D cell culture ...... 8 2.6 In vitro microenvironments ...... 12 CHAPTER 3: OBJECTIVES AND DESIGN CRITERIA ...... 14 3.1 Research goals ...... 14 3.2 Problem identification ...... 14 3.3 Technical design objectives and system design ...... 15 3.4 Design constraints ...... 19 CHAPTER 4: METHODS ...... 20 4.1 Initial design: cell-seeded gels embedded in a collagen matrix ...... 20 4.2 Initial micropatterning experiments to determine substrate, incubation times, and seeding densities ...... 21 4.3 Micropatterned 3D system and a collagen gel overlay ...... 22 4.4.1 Formation of PDMS stamps ...... 22 4.4.2 Preparation of PDMS stamp and substrate for stamping with fibronectin ...... 23 4.4.3 Creation of the micropatterned substrate & cell seeding ...... 24 4.5 Microscopy and data analysis ...... 25 4.6 Micropatterned 3D System with gasket ...... 26 A PDMS gasket was fabricated to physically hold the collagen gel in place (Fig 8). .. 26 4.7 Fabrication of the gasket and treatment with poly-L-lysine...... 26 4.8 Application of the gasket...... 27 4.9 Myofibroblast differentiation ...... 27 CHAPTER 5: RESULTS ...... 29 5.1 Observations of the CSCG system...... 29 5.2 Verification of homogeneous and repeatable micropatterns ...... 29 5.3 Determining optimal seeding density, seeding ratio, and seeding time ...... 30 5.3 Micropatterning, collagen overlay, and live cell microscopy results ...... 32 5.4 Myofibroblast differentiation ...... 40 CHAPTER 6: DISCUSSION ...... 42 6.1 Cell seeded collagen gels within a collagen matrix ...... 42 6.2 Micropattern cell seeding optimization ...... 42 6.4 Myofibroblast differentiation ...... 48 6.5 Design constraints ...... 50

v

CHAPTER 7: SIGNIFICANCE & FUTURE DIRECTIONS ...... 52 LIST OF REFERENCES ...... 55 APPENDIX A ...... 60 MATLAB code to find the distances between tumor and cancer cells and plot the results as a histogram ...... 60 MATLAB code to find the percentage of cells expressing α-SMA ...... 61 MATLAB code to calculate distances between patterns ...... 62

vi

LIST OF FIGURES

1. The components of the tumor microenvironment……..……..……..……… 4 2. Soluble factors secreted by fibroblasts that aid in tumor progression……... 5 3. Differences between 2D and 3D cell culture……..……..……..……..……. 9 4. Current 3D models……..……..……..……..……..……..……..…………... 13 5. 40 µm diameter fibronectin circle stamps……..……..……..……..………. 16 6. Cell seeded gels in a collagen matrix……..……..……..……..……..……... 20 7. Micropatterned system with collagen overlay……..……..……..…………. 22 8. Micropatterned gasket system……..……..……..……..……..……..……... 26 9. The edge of a cell seeded collagen gel showing MDA-MB-231 migration.. 29 10. Circle patterns……..……..……..……..……..…………………………….. 30 11. Patterned fibroblasts and tumor cells co-cultured……..……..……..……… 31 12. Comparison of cell attachment on plastic and PDMS……..……..……….. 32 13. Patterned fibroblasts (red) with tumor cells and only tumor cells………… 32 14. Fibroblasts and cancer cell interaction within collagen gel……..………… 33 15. Live cell microscopy of migrating cell……..……..……..……..…………. 34 16. Cell spread on patterns over time……..……..……..……..……..………… 35 17. Example MATLAB distance quantification……..……..……..……..…….. 36 18. MATLAB distance histogram comparing total number of fibroblasts to cancer cells……..……..……..……..……..……..……..……..……………. 37 19. Histogram from a central pattern to every other pattern……..……..……… 37 20. Histogram of only cancer cell distances over time……..……..…………… 38 21. Histogram comparing individual fibroblasts to cancer cells……..………… 39 22. Myofibroblasts stained for α-SMA……..……..……..……..……..……….. 40 23. Steps to quantify percentage of fibroblasts expressing α-SMA……..……... 41 24. Bar graph showing percentage of α-SMA expressing cells………………... 41

vii

viii

ABSTRACT

The design of a novel 3D micropatterned tumor microenvironment model to study the intercellular effects of fibroblasts and cancer cells

Christine M. Ho Adrian C. Shieh, Ph.D.

The tumor microenvironment contains a multitude of , stromal cells such as fibroblasts, endothelial cells, inflammatory cells, and signaling molecules that support the tumor’s growth and progression. Particularly around the site of tumors, fibroblasts are “primed” to adopt a myofibroblastic phenotype, and are known as cancer associated fibroblasts (CAFs) that promote tumor growth and progression instead of aiding the body. Much interest has been taken in the individual cellular interactions between tumor cells and the surrounding cancer-associated fibroblasts. Systems are constantly being designed to better emulate the complex internal environment around the tumor in order to understand how the tumor stroma communicates and affects the tumor growth. In this thesis, a 3D microenvironment model was developed that simulates the native and allows for the study of individual cellular interactions between cancer cells and fibroblasts, particularly those involving transmission of cell- generated contractile forces. The final system design in brief involves spatially constraining MDA-MB-231 human breast adenocarcinoma cells and human dermal fibroblasts on fibronectin micropatterns on a polydimethylsiloxane (PDMS) substrate. A poly-L-lysine/glutaraldehyde treated PDMS gasket was placed around the cells, and a collagen overlay was added, to allow cells to interact with a 3D environment. Live cell microscopy was used to track cell movement over time, and the images were analyzed with MATLAB. Additionally, a protocol for differentiating myofibroblasts was

ix optimized for the inclusion into the system. The significance of this study is that it offers a novel 3D cell culture system that boasts of the ability for control over individual cell positions in order to observe the effects of fibroblasts and cancer cells in a 3D environment that resembles the in vivo environment more than that of the standard 2D culture system.

x

1

CHAPTER 1: INTRODUCTION

The tumor microenvironment contains a multitude of biomolecules, signaling molecules, and stromal cells such as fibroblasts, endothelial cells, inflammatory cells that support the tumor’s growth [1]. In the vicinity of tumors, fibroblasts are “primed” to adopt a myofibroblastic phenotype, and are known as cancer associated fibroblasts

(CAFs) that promote tumor progression. Myofibroblasts are terminally differentiated cells that express traits characteristic of both smooth muscle cells and fibroblasts, by not only being very contractile, but also continuing to deposit extracellular matrix (ECM).

Much interest has been taken in the individual cellular interactions between tumor cells and the surrounding cancer-associated fibroblasts. Systems are constantly being designed to better emulate the complex internal environment around the tumor in order to understand how the tumor stroma communicates and affects tumor growth [2]. It is known that the highly contractile CAFs generate biomechanical forces that remodel the

ECM and affect the surrounding cells [3]. However, more understanding is needed regarding the interactions between individual fibroblasts, tumor cells, and the ECM, especially in a 3D environment. A 3D system that mimics the in vivo environment is desired because cells behave differently when cultured in 2D environments as opposed to when grown in their native 3D environment [2]. Therefore, the objective of this thesis is to develop a suitable 3D biomimetic system to study the interactions between CAFs and tumor cells.

The specific aims of this project are to (1) study the interactions between fibroblasts and tumor cells through a force mediated mechanism and (2) design a 3D system that will mimic the in vivo tumor environment and allow for the study &

2 quantification of the inter-cellular interactions between fibroblasts and cancer cells. There are several criteria that this system will have to fulfill in order to be deemed successful such as 1) mimicking the mechanical properties of the in vivo ECM 2) allow for study of intercellular effects 3) devising a protocol optimizing differentiation into myofibroblast for inclusion into this system. A method to constrain cells at initial positions to allow for study of individual cell interactions and the quantification of the total number of cells is needed. Constraining cells to known initial positions allows for their movement to be tracked over time to determine if fibroblasts are affecting the movement of tumor cells. Specifically, attention will be given to the polarization and alignment of the cells on the substrate and the collagen gel. Therefore, the deliverable of this thesis is a system that mimics the 3D in vivo microenvironment that allows for the study of intercellular interactions between fibroblasts and tumor cells.

3

CHAPTER 2: BACKGROUND

2.1 Tumor microenvironment

Cancer cells are abnormal cells characterized by their ability to grow and proliferate at an unregulated pace by evading apoptosis inducing factors and growth suppressant hormones due to genetic errors in oncogenes and tumor suppressor genes [CITE].

Frequently, cancer cells have been observed to migrate from the primary site of the tumor through tissues, and metastasize to distant organs. These metastases, rather than primary tumors, are the cause for 90% of all cancer [4]. The specific mechanisms and pathways by which cancer cells spread throughout the body have not been completely elucidated, but much research is being done in this subject. In recent years, scientists have discovered that the environment surrounding the tumor, coined the tumor microenvironment, plays a critical role in supporting the tumor’s growth and promoting its malignancy by a variety of factors. Specifically, the tumor microenvironment consists of a multitude of biomolecules, mechanical forces, and various heterogeneous cell types, including fibroblasts, endothelial cells, and inflammatory cells, that support tumor growth and progression (Fig. 1).

4

Figure 1: Diagram of the tumor microenvironment showing the tumor cells (green), cancer associated fibroblasts (blue stars), immune cells (orange and black circles), and collagen fibers (purple strands) [5] Tumor stroma is remarkably different from the stroma around healthy tissue. This reactive or desmoplastic stroma contains an abnormally high number of activated fibroblasts, infiltrating inflammatory and immune cells, in addition to exhibiting increased matrix stiffness [6], aberrant angiogenesis [7], and elevated interstitial fluid pressure [8]. The components of the tumor stroma resembles granulation tissue seen in wounds, giving tumors the moniker, “wounds that do not heal” [9]. Furthermore, cytokines and proteases secreted by stromal cells and tumor cells, such as vascular endothelial growth factor (VEGF), matrix metalloproteinases (MMPs), and transforming growth factor beta (TGF-ß), directly and indirectly alter the ECM. As a result, the surrounding activated stroma continuously provides support for tumor growth and progression.

2.2 Cancer associated fibroblasts (CAFs)

Fibroblasts are the most prevalent cell type found in the of the body. The main functions of fibroblasts are to synthesize new ECM, regulate

5 inflammation and epithelial differentiation, and assist in wound healing by contracting the surrounding tissue to minimize the wound area [10]. However, in the presence of tumors, normal fibroblasts are initially “primed,” then activated to become cancer- associated fibroblasts (CAFs) which contribute to the tumor’s growth, proliferation, and migration by chemical secretions and physical activities [9]. For example, CAFs secrete matrix metalloproteinases (MMPs), transforming growth factors (TGF), collagen, and fibronectin that remodel and degrade the ECM to become more permissive to tumor migration (Fig 2) [10]. As fibroblasts migrate toward the site of the tumor, they release

TGF-ß which induces additional fibroblast proliferation, recruitment, differentiation into the -SMA expressing myofibroblast cell type.

Figure 2: Cancer associated fibroblasts secretes various factors to aid in tumor progression[10]

Myofibroblasts are present at all stages of tumor growth and share characteristics of both fibroblasts and smooth muscle cells [11]. In particular, myofibroblasts are highly contractile cells that not only exert mechanical forces on the microenvironment by constant pulling and contracting but also continuously depositing ECM and secreting cytokines. Through the application of biomechanical forces and the chemical secretion of

6 growth factors, cytokines, and collagen, CAFs remodel the tumor stroma by both physical and chemical means into a more tumor-friendly environment that promote tumor growth and progression.

2.3 Effects of soluble factors on the tumor environment

Chemical secretions of various cytokines and growth factors by stromal fibroblasts induce ECM remodeling. Stromal fibroblasts are the primary sources for tumor-promoting TGF-ß, VEGF, and MMPs in the tumor microenvironment. The cytokine TGF-ß has both tumor-suppressing and tumor-promoting effects, which depend on the stage of carcinogenesis. Normally, TGF-ß mediates the activation of fibroblasts into -SMA expressing myofibroblasts during wound healing and fibrosis.

However, the TGF-ß secreted by the increased number of CAFs in the tumor microenvironment causes autologous activation of fibroblasts and induces CAF differentiation in a positive feedback loop. TGF-ß acts as a tumor-promoting factor by recruiting fibroblasts and inducing epithelial-mesenchymal transition (EMT) of cancer cells [12]. Cancer cells undergoing EMT become more migratory, thus increasing the chance of metastasis. Migratory cancer cells typically travel through the lymphatic system and the circulatory system to distant sites. Not surprisingly, abnormal lymphangiogenesis and angiogenesis induced by secreted VEGF occurs in the tumor vicinity and further contributes to tumor growth and metastasis.

VEGF is a growth factor that is in involved in angiogenesis, vasculogenesis, and lymphangiogenesis. There are four types of VEGF: VEGF-A, VEGF-B, VEGF-C, and

VEGF-D, however VEGF-D and VEGF-A are the subsets most commonly involved in angiogenesis and lymphangiogenesis in the tumor microenvironment [13]. VEGF-D

7 induces both angiogenesis and lymphatic formation within tumors, leading to metastasis via tumor cell dissemination through lymph nodes, while VEGF-A only induces angiogenesis [13]. The formation of new lymphatic tissue and blood vessels provide tumor cells a gateway to migrate away from the primary tumor site and spread to distant organs. The fluid flow in vasculature and lymphatic tissue also assist in sustaining the tumor’s growth and progression through diffusion of nutrients, and waste removal.

MMPs contribute to the high turnover rate of the ECM due to their matrix degrading properties. MMP-mediated ECM degradation fosters tumor progression and cancer cell invasion. Furthermore, MMPs also interfere with the induction of cancer cell apoptosis by inhibiting receptor-transmitted apoptosis, leading to increased size of tumors

[14].

2.4 Mechanical effects on the tumor environment

The dynamic ECM microenvironment is susceptible to modification and remodeling due to various physical factors, in part due to the mechanical forces produced by contractile fibroblasts.

Fibroblast contractions are carried out by the motor-protein myosin II, which is responsible for the actin-based motility of cells. Myosin II is largely regulated by two mechanisms: myosin light chain kinase (MLCK) and the Rho-associated kinase (ROCK)

[4]. Cell contractility does not only function in remodeling the ECM, but also has a role in modulating cell migration and invasion [15]. Due to the contractile CAFs constantly pulling and applying traction forces on the microenvironment, studies have shown that these mechanical forces reorganize the collagen fibrils within the ECM [16]. These traction forces remodel the ECM by reorganizing the randomly orientated collagen fibers

8 into parallel networks, causing them to become more permissive to tumor cell migration

[11]. The parallel collagen fibers act as “tracks” that are hypothesized to facilitate tumor cell invasion [17]. A study by Provenzano et al. showed that matrix organization through a force-mediated mechanism influences cell migration in a 3D environment. Two small tumor cell seeded collagen gels (CSCG) were added into a collagen matrix with 7 mm between the CSCG. Tumor cells were observed to exert mechanical forces that aligned collagen fibrils in the matrix and subsequently migrate along these fibrils [18]. Less cell migration was present in regions with low fibrillar collagen alignment, highlighting the importance of the effects of matrix organization on cell migration.

In addition to matrix organization, the mechanical and material properties of substrates are also influential in cellular communication and responses. The interactions of cells in non-linear elastic substrates were studied by Winer & Jamney. Fibroblasts were grown in 3D collagen matrices and observed to be able to remodel the collagen fibrils up to five cell lengths away [19]. The production of local stiffness gradients allowed cells to sense and respond to other cells though a force limited mechanism up to hundreds of microns away in non-linear anisotropic matrices. This study brought to light the acute responsiveness of cells to local substrate deformation and ability to induce local stiffness gradients that can affect neighboring cells hundreds of microns away.

2.5 Comparison of 2D and 3D cell culture

Cell culture has traditionally been performed on 2D plastic or glass substrates.

Even though growing cells in 2D environments allows for relatively straightforward culture and observation, cells cultured on 2D substrates express drastically different behaviors due to markedly different growth conditions than in a 3D environment [20].

9

Culture in the highly simplified 2D environment limits the native cell-cell communication methods, alters cellular morphology, and changes both mechanical and biochemical cues [21]. The rigidity and stiffness of the substrate affects the cellular morphology (Fig 3), spreading, and the direction of movement and migration. In contrast to rigid non-deformable glass or plastic substrates with elastic moduli ~1GPa [6], cells in vivo usually bind to ECM with elastic moduli ranging in the scale of 10 - 10,000 Pa [22].

Figure 3: Fibroblasts spread and flatten more on hard, 2D tissue culture plastic (left) and are more spindle-like in a softer, 3D environment [4]

It is biologically important for cells to be able to sense and respond to their environment. Fibroblasts are one example of a cell type that presents morphological differences when cultured in 2D and 3D substrates with varying stiffness. Specifically, the substrate’s mechanical stiffness affects the cell’s spread and geometry, which are two characteristics that affect cell differentiation, proliferation, and even apoptosis [23].

Membrane bound focal adhesion receptors first bind to the substrate to allow for initial cell attachment and subsequent sensing of the matrix stiffness by the pulling of the cytoskeleton. [24], [25]. For example, fibroblasts are contractile cells which remodel their surrounding environment by constant pulling and contracting. However, stiff substrates such as tissue culture plastic do not deform in response to the fibroblasts’ applied forces.

Therefore, fibroblasts have been observed to spread out and adopt a flattened morphology, without articulated stress fibers on stiff substrates [26]. In soft connective

10 tissues, fibroblasts do not spread and flatten, but adopt a spindle (stellate) shape with round nuclei [27]. Substrate stiffness not only affects the geometric shape of cells, but also cellular migration.

By adhering to the tissue culture plastic, only one side of the cell is attached to the substrate, while the other side is exposed to the culture medium. This causes a polarization in integrin binding and mechanotransduction leading to atypical interactions with soluble biomolecules and unnatural migration. 2D and 3D environments affect the cell’s migration speed, persistence (the distance and time the cell travels before it changes directions), and type of migration (mesenchymal or amoeboid) [28]. In 2D environments, cell migrational speed depends on the strength of the adhesion complexes to the substrate, whereas migration speed in 3D is affected by the mechanical and structural properties of the matrix [29].

One example of how the mechanical properties of the substrate affect cells is the stiffness of the substrate. Cells sense substrate rigidity and migrate accordingly.

Specifically, cells carry out the phenomenon of durotaxis, or directed movement along the stiffness gradient of the substrate [26]. The migration of individual cells to a stiffer substrate and rejection of a more compliant substrate was observed with 3T3 fibroblasts.

Specifically, in a study done by Lo et al., 3T3 fibroblasts were seeded on collagen-coated polyacrylamide sheet with varying stiffness. Fibroblasts tactilely sensed substrate stiffness by reaching out and actively exploring the environment. It was observed that the fibroblasts preferentially migrated towards stiffer substrate regions, and pause and/or reverse at the boundary when migrating from a stiffer region to a less stiff region [30].

Since cells are acutely aware of the rigidity of their environment and accumulate on

11 regions with higher stiffness [31], consideration of the mechanical properties of the substrate should be taken when culturing cells for study.

In addition to the effects of the substrate stiffness, the components of the native tissue should be considered. The hard, flat isotropic surface of tissue culture plastic differs greatly from the softer, anisotropic three-dimensional native ECM architecture that envelops cells and provides structural support in the body. Cells thrive in the soft 3D stroma, which is largely made up of the proteins fibronectin and collagen types I and III.

Collagen is the most prevalent protein in the body and is found within the connective tissue and the extracellular matrix as a meshwork of collagen fibrils, offering mechanical support for the ECM [32].At least 16 types of collagen have been discovered, but collagen types I, II, and III comprises 80-90% of all the bodily collagen. Type I collagen is prevalent in the tumor stroma, , skin, , ligament, and interstitial tissues [32].

Many collagen fibrils pack together into long collagen fibers that provide structural support in the tumor microenvironment. The collagen density within the vicinity of tumors affects tumorigenesis and cancer cell migration, and metastasis. For example, the risk of more malignant tumors is higher in patients with increased stromal collagen in breast tissue [33]. This increased change of malignant tumors could be due to the mechanical signals and adhesion mediated signals that stromal collagen provides on the tumor cells [34].

Fibronectin is another common component in the ECM that exists as a high- molecular weight, multimodular on the order of ~440k kDa. Fibronectin mediates various processes involving adhesion, cell-cell and cell-matrix interactions, growth, differentiation, and migration. Within the tumor stroma, fibronectin secreted by

12 fibroblasts is linked to angiogenesis [35] and serves as a chemoattractant for fibroblasts[36]. Fibronectin is most notably known for its role in cell adhesion. Integrins, which are cell membrane receptor proteins, interact with the binding affinity sequences on the fibronectin, collagen, and other matrix proteins to allow for adhesion. Fibronectin functions as a ligand for many membrane-spanning integrin receptors such as its primary receptor, integrin α5β1. Fibronectin also expresses the well-known arginylglycylaspartic

(RGD) acid tripeptide binding sequence comprised of L-arginine, glycine, and L-aspartic acid, active during integrin-dependent cell attachment and cellular recognition.

2.6 In vitro microenvironments

The complexities of the 3D in vivo ECM are extremely difficult to create and analyze during cell culture due to the intricate interplay of host cells, cytokines, protein fibers, fluid flow, and mechanical forces. Due to the simplicity of 2D culture systems and the abnormal cellular responses, there is a great need to develop a better 3D system to mimic that of the in vivo environment to allow for cellular observation. Even though there will still be disparities between the 3D system and the complicated in vivo environment, a new 3D culture system will help bridge the gap between the 2D culture conditions and the native tissue. Much interest has been taken in the cell-cell and cell- stroma interactions between tumor cells and surrounding cancer-associated fibroblasts, and how these interactions affect tumor cell migration. Many attempts have been made to design a system to better mimic the complex in vivo tumor microenvironment in order to understand how the tumor stroma communicates and affects tumor growth. To date, there has not been any one system that completely emulates the intricate, complex network of

13 the ECM [2]. However, much work and many system designs have been developed to approach this modeling problem.

Currently there are several types of 3D models that aim to simulate the native tissue to study cell behavior [37]. One type involves growing multicellular cancer spheroids suspended in a protein matrix, such as collagen or Matrigel (Fig 4). Another system models cancer migration by coculturing cancer and fibroblasts within a collagen matrix.

Fibroblasts are embedded in a collagen matrix, while cancer cells are grown in a monolayer above. However, these models lack the option of spatially constraining cells at initial positions in order to study the intercellular effects over time.

Figure 4: (L) 3D model of cancer spheroids in a collagen matrix. (R) 3D model of migration

Therefore, a better 3D system that mimics the in vivo environment is needed in order to elucidate the interactions between individual fibroblasts, tumor cells, and the ECM.

Therefore, the purpose of this thesis is to develop a 3D system to allow for the controlled study of single cell interactions between fibroblasts and tumor cells that mimics the in vivo environment.

14

CHAPTER 3: OBJECTIVES AND DESIGN CRITERIA

3.1 Research goals

The main research goal of this thesis is to study the effects of fibroblasts on tumor cells though a force mediated mechanism. To explore this central question, a novel 3D system needs to be developed to mimic the in vivo microenvironment. This system will allow for the study and quantification of the interactions between fibroblasts and tumor cells with the hope to elucidate the effects of fibroblasts and myofibroblasts on tumor migration in a 3D matrix. This 3D system will offer flexibility to easily alter experimental parameters for future studies. Furthermore, for this system to be deemed successful, it must meet certain design criteria: 1) The device must allow for study of individual cells and intercellular effects between tumor cells and fibroblasts. 2) The system must mimic the tumor environment. 3) A protocol optimizing the differentiations conditions of fibroblasts into myofibroblasts for inclusion into this system must be developed.

3.2 Problem identification

There is a lack of a 3D model that sufficiently mimics the in vivo microenvironment and allows for the study of intercellular effects between fibroblasts and cancer cells at spatially controlled positions. Therefore, the goal of this thesis is to design a controllable, biomimetic 3D system that, in conjunction with a collection of technical methods and computational techniques, will allow us to evaluate and quantify the effects of fibroblasts on tumor cell migration.

15

3.3 Technical design objectives and system design

In order to successfully evaluate and understand the interactions between individual cancer-associated fibroblasts and tumor cells, a system has to be designed that meets specific criteria and allows for measureable results (Table 1).

Table 1: List of the design criteria and specifications of the system Design criteria Design specifications

 The device must allow for study of  Cells must adhere to at least 70% individual cells and the intercellular of the patterns effects between tumor and fibroblasts. The total amount of cells  There should be only one cell on and total time for cell seeding to 70% of the total number of patterns allow for spatial constraint must be with cells optimized.  Cells must be at a set initial starting position

 The system must mimic the tumor  Collagen gel has to be within a environment range of elastic moduli, 4-12 kPa, to mimic tumor environment [38]

 Must have a 3D aspect

 A protocol optimizing the  At least 60% of the fibroblasts differentiations conditions of should express α-SMA. fibroblasts into myofibroblasts for inclusion into this system must be  A concentration of TGF should be developed. determined and be deemed statistically significant from the other conditions.

The first device criterion must allow for the study of individual cells at specified initial positions and their resulting intercellular effects. In order to fulfill this criterion, the design specification of having at least 70% of cells in a field of view spatially

16 constrained must be achieved. In order to observe the extent of local intercellular effects between tumor cells and fibroblasts at regulated distances, a technique to constrain individual cells at specific distances and orientations is needed. Micropatterning is a technique that stamps geometric patterns in the micro- or nanoscale that will allow for distance control between cells. Specifically, micropatterning, or microcontact printing, will spatially organize cells at specific distances and allow for the study their intercellular interactions [23]. An adhesive protein will be patterned on to a substrate to allow cells to uptake the patterns and specific spatial distances. Since cells express a high affinity to fibronectin, fibronectin will be chosen as the “ink” for the patterns. The physical shape of cells influences their behavior in many cellular processes including differentiation, apoptosis, motility, and proliferation [39]. In this system, micropatterning will be used to isolate each cell at a specified distance in order to observe physical intercellular effects such as alignment, movement toward each other, or extension of filopodia. A PDMS stamp will be fabricated with a test pattern of 40 µm circle islands with 100 µm between each island (Fig 5). The shape, size, orientation, and spacing of the patterns are a customizable aspect the system.

Figure 5: 40um diameter circles of fibronectin were stamped on to a slightly hydrophilic PDMS-coated 6- well plate.

Cells will adhere to the fibronectin islands and take up the said geometric shapes at predefined initial positions. The design criteria of constraining cells on individual patterns can be quantified by calculating the percentage of patterns that are taken up by

17 cells and how many patterns have more than two cells on them. Specifically, the design specification of having at least 70% of the micropatterns with adhered cells must be met.

Having at least 70% of micropatterns filled with cells will allow there to be enough cells spatially constrained for study. Furthermore, of these micropatterns filled with cells, there should only be one cell per pattern for at least 70% of all the cells already on the patterns.

Since interest mainly lies in the individual cell behavior, having more than one cell on patterns is not desired. In order to achieve the desired 70% of micropatterns filled, various factors must be first optimized. In particular, the seeding concentration of cells and duration of time cells are allowed to uptake the micropatterns must be validated.

The next design criterion is that the system must provide similar chemical and biomechanical characteristics as the nonlinear, anisotropic properties of the vivo microenvironment as well as offer a 3D aspect. To simulate the tumor microenvironment, a 3D ECM mimic will be introduced in the form of a collagen gel overlay on top of the micropatterned cells. Collagen was chosen because it is the most prevalent ECM protein and exhibits nonlinear and anisotropic mechanical properties similar to the ECM. By introducing the collagen overlay, cells will interact with a 3D collagen environment, similar to the one in vivo. The shear modulus, elastic modulus, and mechanical properties of the collagen gel will be scaled to best simulate the in vivo properties. Specifically, the shear modulus of breast tissue is experimentally found to be between 2.0-66 kPa for glandular tissue and 0.5-25 kPa for adipose breast tissue [40]. Studies have shown that the elastic moduli of collagen gels vary with the collagen concentration [41]. The elastic modulus of 2.0 mg/mL collagen gel was found to be between 1-28 kPa in a tensile test

[42]. The elastic modulus of breast cancer tissue is higher than that of normal breast

18 tissue, at 4-12 kPa and 0.4-2 kPa, respectively [38]. Since there is an overlap between the mechanical properties of native tissue, tumor tissue, and reconstituted collagen gels, a range of collagen gel concentrations can theoretically be used. However, for this study, a test concentration of 2.0mg/mL collagen gel will be used. By adding a third dimension, the system does have an additional level of complexity, but allows for observations of intercellular actions in an environment that better simulates the native ECM.

The last design criterion is to allow for the inclusion of not only fibroblasts but of myofibroblasts. Therefore, a protocol optimizing the differentiation conditions of fibroblasts to myofibroblasts must be devised. Specifically, at least 60% of fibroblasts must differentiate into myofibroblasts. In the vicinity of tumors, fibroblasts begin to express -SMA, which is a trademark of myofibroblasts. Since myofibroblasts are more contractile cells and exert higher forces on the local microenvironment, a combination of fibroblasts and myofibroblasts is desired to add to the system.

Through the use of this novel system, the research objectives of observing and quantifying how fibroblasts and tumor cells react when cocultured together, cultured separately, and analyzing the differences between tumor and fibroblasts cultured on 2D tissue culture plates in contrast to the 3D system will be compared. After the collagen gel is added to the system, the cells will interact with the gel and the interplay between fibroblasts and tumor cells will be of specific interest. To determine if fibroblasts are affecting the movement of tumor cells, the distances between the tumor and fibroblasts will be calculated at various time points after initial cell seeding. Attention will be given to any cytoskeletal projections signaling the direction of cell migration as well as the polarization and alignment of the cells on the substrate and the collagen gel. Time-lapse

19 microscopy images will be analyzed by MATLAB in order to identify each individual cell and the changing distances between them.

3.4 Design constraints

Several design constraints need to be considered as potential limitations for this system. Optimal cell attachment to the patterns and reduction of non-specific binding is one hurdle that will affect the design criteria of constraining cells at specific initial positions. Cells need to attach at spatially constrained points because initial cell positions are necessary to track the behavior of particular cells over time and to fulfill the goal of observing intercellular effects in this system.

Another constraint is the availability of equipment and the necessity to integrate the system with the specific Leica fluorescence microscope readily available. The system design must be tailored to be able to acquire data from a fluorescence microscope with only four light channels. Thought must be taken in choosing the type of fluorescent dyes and markers that are compatible with the cells and the imaging capabilities of the microscope. The last constraint is supporting cell viability and preventing cell .

Many factors can results in cell death, such as pH of the media changing, cells drying out, and traces of cytotoxic substances being in the system. For this experiment to be successful, cell toxicity and death must be kept to a minimal.

20

CHAPTER 4: METHODS

4.1 Initial design: cell-seeded collagen gels embedded in a collagen matrix

The first method attempted to study the interactions between tumor and fibroblasts in a 3D environment was modified from the study conducted by Provenzano, et al. 2008. Either the invasive human breast adenocarcinoma cell line, MDA-MB-231, or human mammary fibroblasts were suspended within a 1.0 mg/mL rat-tail collagen mixture, termed the cell-seeded collagen gel (CSCG). CSGCs were formed by dispensing

200 uL of a cell/gel mixture into a 96-well plate, and allowed to polymerize for at least 1 hour at 37°C. A second acellular 2.0 mg/mL collagen gel was allowed to polymerize for

15 min at room temperature. This 2.0 mg/mL gel functioned as the acellular collagen matrix. Two CSCG were transferred into the acellular collagen matrix at a distance of 7 mm apart edge to edge (Fig 6). The CSCGs sunk into the acellular collagen matrix and were incubated at 37°C with 5% CO2. After 2 hours, DMEM media with 10% serum and

1% penicillin streptomycin was added to the assay, and changed every 3 days for the duration of the assay.

Figure 6: 1.0 mg/mL CSCGs (grey) were placed 7mm away in a 2.0mg/mL collagen matrix (blue) in two different geometric alignments.

The assay was run for 3-5 days, then fixed with 4% paraformaldehyde (PFA), and stained with phalloidin and DAPI. A Leica fluorescent microscope was used to analyze the positions and interactions between the cells. Several conditions were compared: 1) adding two tumor CSCGs, 2) adding two fibroblast CSCGs, 3) adding one tumor and one

21 fibroblast CSCG, and 4) adding two tumor CSCGs and one fibroblast CSCG in a triangular pattern.

4.2 Initial micropatterning experiments to determine substrate, incubation times, and seeding densities

Preliminary experiments to determine length of time to fibronectin on the PDMS stamp, the length of stamping time, the type of substrate with best stamping. First, fibronectin was incubated on the PDMS stamp for 30 mins, 1 hour, 1.5 hour and 2 hours.

The stamping time varied from 10 mins, 30 mins, and 1 hour on either glass or tissue culture plastic substrate. To block cell attachment to the unstamped substrate, hydrophilic

PEG/PLL or Pluronic F127 was adsorbed onto the substrate. Next, the cell seeding density and the time until proper cell adhesion was determined. Various cell seeding densities were attempted to allow for enough cells to attach to the fibronectin patterns, but not too much to increase the chance for more than one cell attaching to the same pattern. Cells were seeded at 40,000, 60,000, 80,000, 100,000, 150,000, and 200,000 cells in each well of a six well plate. An optimal ratio between the tumor cells and fibroblasts were also determined. Ratios of 1:1, 1:2, 1:3, 1:5, 1:7, and 1:10 of fibroblasts to tumor cells were tried. The amount of time it took for cells to adhere and start spreading out on the substrate was noted.

22

4.3 Micropatterned 3D system and a collagen gel overlay

A 3D system was designed for the study of the single cell interactions between fibroblasts and tumor cells at initial positions predefined by stamping fibronectin islands in the process of micropatterning (Fig 7).

c a b

Figure 7: Proposed fibronectin micropatterned system with collagen overlay. a) A PDMS stamp (top, brown) is used to stamp fibronectin patterns (purple). b) MDA-MB-231 (red) and fibroblasts (green) are seeded on to the patterns. c) A collagen gel overlay (blue) is applied.

4.4.1 Formation of PDMS stamps

PDMS stamps were synthesized through the process of photolithography. A negative master mold made from silicon wafers and SU-8 photoresists, kindly borrowed from the Clyne lab, was used to cast the polydimethylsiloxane (PDMS) stamp. The silicon master was etched with a grid of 100 by 100 40 µm diameter circles. The master, along with a few drops of tridecafluoro-1,1,2,2-tetrahydrooctyl-1-trichlorosilane

(TFOCS), were placed into a vacuum desiccator for 30 minutes. The TFOCS was vaporized and deposited as a monolayer onto the master through siloxane bonding, transforming the hydrophilic surface of the master to become hydrophobic. Thus, the

TFOCS assisted in the cured PDMS stamp release from the master by preventing it from sticking to the silicon wafer. PDMS was made by thoroughly mixing a silicone elastomer

23 base and a Sylgard 184 elastomer curing agent (Dow Corning) at a 10 to 1 ratio. The

PDMS mixture was poured onto the master and placed into a vacuum dessicator to remove bubbles. The PDMS was cured at 70°C for 2 hours, then carefully peeled off the master and the excess PDMS was trimmed to form the stamp.

4.4.2 Preparation of PDMS stamp and substrate for stamping with fibronectin

The bottoms of 6-well plates were coated with approximately 0.8g of PDMS

(enough to cover each well), and was cured at 60°C for 3 hours. The wells were sterilized by UV light (365nm) for 15 mins. Since PDMS is too hydrophobic to allow for protein absorption, it was treated by UV/ozone to activate the surface and render it slightly less hydrophobic. Due to the phenomena of migration of low molecular weight uncrosslinked polymeric chains to the surface from the bulk surface, the modified surface will regain its original hydrophobocity in just several hours. Therefore, to ensure the functionality of the

PDMS substrate, it was stamped within 30 mins from the UV/ozone treatment.

Stamps were cleaned by sonicating for 5 mins in 2% sodium dodecyl sulfate

(SDS), sterilized by submerging in ethanol for 5 mins, and washed with sterile DI water.

50µl of 50µg/mL bovine fibronectin (Sigma-Aldrich) diluted in DI water or 1X PBS was incubated on PDMS stamps for 1 hour. The stamp was washed with three changes of 1X

PBS for 2 mins each to remove excess fibronectin, and allowed to fully dry before stamping. In order to stamp the fibronectin patterns, the PDMS stamps were careful pressed on to the PDMS-coated wells and left for 15 sec to 1 min. After stamping, the stamps were cleaned by putting stamp side up, then adding DI water until it covers the stamps to remove the fibronectin protein residue. Then, stamps were rinsed a second time in DI water.

24

4.4.3 Creation of the micropatterned substrate & cell seeding

The PDMS coating was blocked with either 0.2% or 0.02% Pluronic F127

(diluted in 1X PBS) for 30 mins in order to prevent cells from adhering to the unstamped regions of the PDMS substrate. Therefore, cells could only attach to the patterned fibronectin islands. The PDMS was then washed three times with 1X PBS to remove residual Pluronic F127.

MDA-MB-231 and dermal fibroblasts between passages of 5-19 were cultured in

Dulbecco’s Modified Eagle Medium with 10% serum and 1% penicillin streptomycin. In order to differentiate the cell types, CellTracker Blue CMAC (15uM), Red (10uM), or

Green (10uM) (Invitrogen) were used. Cells were incubated for 30 mins with CellTracker dyes in serum free media to facilitate uptake, then allowed to recuperate for at least 30 mins in full DMEM media with 10% serum and 1% penicillin streptomycin. Cells were seeded at a total cell density of 100,000 cells/well in the 6-well plate; the ratio of fibroblasts to tumor cells tested varied from 1:2 to 1:10. The media and cells were rehomogenized after 30mins, then after 1 hour, the media was aspirated out, and the cells were were washed once with 1X sterile PBS. Full DMEM media was added, and cells were incubated for 3 hours at 37°C and allowed to take up the patterns.

After 3 hours, the media was removed, cells were washed once with 1X sterile

PBS, and a 2.0 mg/mL collagen was applied on top of the cells. The collagen gel was allowed to polymerize at 37°C for 1 hour, then DMEM media was added.

25

4.5 Microscopy and data analysis

Brightfield and fluorescent images were taken just before collagen gel application

(Time 0), and at time points 2 hours (Time 1), 5 hours (Time 2), 14 hours (Time 3), and

26 hours (Time 4) after gel application 5 hours (Time 5). Cells were tracked by a displacement method based on a Cartesian grid. The bottom of the plate was marked with a marker. The mark was found under the microscope and given the coordinates (0,0). The coordinates of the desired cells were marked and used to refind them at each time point.

Furthermore, live cell microscopy was performed after the collagen gel was applied. DMEM media was removed from the assay in favor of Leibovitz’s L15 media to allow imaging in conditions without 5% CO2 to prevent a change of DMEM media pH.

Brightfield and fluorescent images were taken every 15 mins for ten hours. Cells were then fixed with 4% paraformaldehyde for 1 hour, washed thrice with 1X PBST, and stained with Phalloidin 488 (1:50) and DAPI (1:500)

MATLAB and the Image processing toolbox were used for image analysis of the fluorescent and bright field images. A code was developed to isolate the two types of cells based on the microscope channels, with the goal to calculate the distance away from individual cells over a time period. Specifically, images were denoised and thresholded using the adaptive Otsu method. Blob segmentation was performed to isolate individual cells by applying the watershed algorithm and command bwlabel to count each cell. The distances between each cell were calculated by using the positions given by regionprops.

Cell circularity and area were calculated. Circularity was defined as the , where P

is the perimeter and A is the area of the cell.

26

4.6 Micropatterned 3D System with gasket

A PDMS gasket was fabricated to physically hold the collagen gel in place (Fig 8).

a b c d

Figure 8: The fibronectin micropattern system with a PDMS gasket placed over the micropatterned cells. a) Stamping fibronectin patterns with a PDMS stamp b) Seeding cell c) Adding a PLL-glutaraldehyde treated PDMS gasket to crosslink with the collagen d) Adding the collagen overlay in the gasket

4.7 Fabrication of the gasket and treatment with poly-L-lysine

A gasket was prepared from PDMS at a 1:20 crosslinker ratio and treated with poly-L-lysine and glutaraldehyde in order to crosslink with the collagen overlay and hold it in place over the patterned cells, as explained below. A1x1cm square hole large enough to expose the fibronectin stamped area was cut in the center of each PDMS gasket. The gasket was plasma-treated at full power for 1 min. The gasket was placed on parafilm and incubated in 1% (0.1 mg/mL) of poly-L-lysine for one hour, rinsed three times with DI water, then allowed to completely dry. Then, the gasket was incubated in 1% glutaraldehyde for 1.5 hours under the hood, and rinsed with three times with DI water.

The gasket surface was coated with a 50 µg/mL collagen solution diluted in PBS for 24 hours in order to leach out excess glutaraldehyde. Care was taken to only surface modify the inside walls of the gasket to reduce excess exposure to glutaraldehyde. Sterilization was accomplished by placing under UV (at 365nm) for 10 mins, then washed with 70% ethanol for 5 mins, and rinsed with water.

27

4.8 Application of the gasket

Cells were seeded on to the patterns and allowed to spread for 2 hours before media was aspirated, and washed once with 1X PBS. The treated gasket was place around the patterned area on the PDMS substrate, and adhered to the substrate via covalent bonding between the gasket and substrate. A 2.0 mg/mL collagen gel was added inside the gasket and allowed to polymerize at 37˚C for 1 hour. After 1 hour, L15 media was added to the well. Live cell microscopy was taken and images were processed with

MATLAB. A MATLAB code was developed to isolate each cell and to find the distance from a single fibroblasts to each cancer cell (Appendix A). The first step was to separate the image channels in order to differentiate the cell lines. Then, the images were thresholded, denoised, and counted by the methods described above. The distances between each cell was calculated from centroid to centroid. Each fibroblast was isolated, and the distance between its centroid to the centroid of every cancer cell was calculated until all the fibroblasts were analyzed.

4.9 Myofibroblast differentiation

The optimal protocol for differentiating human mammary and dermal fibroblasts was determined by testing numerous experimental parameters such as varying concentration of TGF-ß[43], concentrations of serum[44], adding a serum starvation time period[45], length of time incubated with TGF-ß [46], and number of additional TGF-ß changes during the culture time [47].

Concentrations of 1, 2, 5, 10, 15, and 20 ng/mL TGF-ß were attempted, along with variations in serum concentration ranging from no serum, 1%, and 10% serum, adding a serum starved period for a day or no serum starved period, and incubating cells

28 with TGF-ß for 3,4, or 7 days. Fibroblasts were cultured in 10% serum media with 1% penicillin streptomycin and split every 3-4 days. 10,000 fibroblasts were seeded into a

12-well plate, or 7,000 cells into an 8-chamber glass bottomed slide. The cells were incubated for 24 hrs in full DMEM media with 10% serum and 1% penicillin streptomyocin. Then, each concentration of TGF-ß was diluted into serum free, low serum (1%), or high serum (10%) media and added to the cells. Fibroblasts were cultured in TGF-ß for 3, 4, or 7 days, changing media every 3 days. Cells were fixed with 4% paraformaldehyde and stained with Phalloidin 488, DAPI, and α-SMA. Fluorescent images were taken with a Leica microscope, and a MATLAB code was written to analyze the expression of α-SMA and the myofibroblastic phenotype. Specifically, the images were split into three channels, one for each stain. Images were denoised and thresholded using the adaptive Otsu method. Blob segmentation was performed to isolate individual cells by applying the watershed algorithm and command bwlabel. The DAPI channel was analyzed to get the total number of cells. The α-SMA channel was isolated and the fluorescent intensity was calculated for each cell expressing α-SMA. Cells expressing α-

SMA were counted and the percentage of differentiated myofibroblasts out of the original fibroblasts was calculated. The conditions that resulted with the greatest percentage of myofibroblasts were selected for future differentiation experiments.

29

CHAPTER 5: RESULTS

5.1 Observations of the CSCG system

Fibroblasts and tumor cells were observed to branch out and migrate from the initial CSCGs into the accellular 1.0mg/mL collagen matrix (Fig 9). Furthermore, small colonies of cancer cells were observed to grow sporadically in the acellular matrix. There was lack of spatial control over the cells in the gels and the intercellular effects between the fibroblasts and the tumor cells were not easily quantifiable due to the large number of cells.

Figure 9: MDA-MB-231 cells were observed to migrate into the collagen matrix (20x). Scale bar represents 100 μm.

5.2 Verification of homogeneous and repeatable micropatterns

First, micropatterning was attempted on glass bottomed slides and tissue culture plastic. When stamped on glass and plastic substrates, the fibronectin was incubated on the stamp for 1.5h, then stamped on the substrate for 45mins. However, no workable stamps were made on the glass-bottomed slides, and the tissue culture plastic stamps were not homogenous nor easily repeatable. However, when stamped on an oxygen

30 plasma treated PDMS substrate, homogenous micropatterns of fibronectin at specified distances of 40 um diameter circles, 100 um apart, were successfully achieved. For these experiments with the oxygen plasma treated PDMS substrates, the fibronectin was incubated on the stamp for 1h and the PDMS stamp was placed on the PDMS substrate for 1 min (Fig 10).

Figure 10: 40 um diameter circle stamps of 50 ng/mL fibronectin (20x) on (a) tissue culture plastic and (b) oxygen-plasma treated PDMS. Scale bar represents 100μm.

5.3 Determining optimal seeding density, seeding ratio, and seeding time

The optimal substrate to support cell viability as well as fibronectin adhesion was first determined. Tissue culture plastic and glass bottomed slides were chosen because the former are frequently used substrates for cell growth and the latter offers a better imaging substrate for microscopy. Cells were observed to grow on both tissue culture treated polystyrene plates and glass bottom plates. However, the fibronectin patterns were unable to be formed on the glass substrate. Patterns on the plastic surface were not reliably reproducible.

Optimal seeding density was determined to be 100,000 total cells to allow for cells to attach to patterns. Next, the time it took for each cell line to adhere to the substrate was determined to be 1 hour for the cancer cells and 30 mins for the fibroblasts.

Optimal time for cell seeding was determined by visual inspection of the cells every 10

31 mins under a light microscope to observe initial attachment and then subsequent cell spreading on the substrate. The morphology of cells suspended in media and those attached to the substrate is different. For example, cells suspended in media adopt a round, ball shape, while cells adhering on the substrate become flatter. It was noticed that the MDA-MB-231 cell line took longer to adhere to the patterns than the dermal fibroblasts. When seeding both cells types at the same time and washing off the excess cells after 1 hour, most of the cells adhered to the micropatterns were fibroblasts.

Therefore, to allow tumor cells a chance to adhere to the patterns, dermal fibroblasts were seeded 45 mins after MDA-MB-231 cells were added, and left to attach for an additional

30 mins. After the cell seeding density and the time for attachment were determined, fibroblasts and cancer cells were seeded into the micropatterned system and successfully took up the circle patterns (Fig 11). The number of cells that took up fibronectin patterns was significantly greater on the oxygen-plasma treated PDMS than on the tissue culture plastic, with a p < 0.05 (Fig 12).

Figure 11: Dermal fibroblasts (red) and MDA-MB-231 cells successfully adhered to the patterns. (20x) Scale bar is 50µm.

32

Figure 12: Significantly more cells took up patterns on oxygen-plasma treated PDMS than on tissue culture plastic (p<0.05)

5.3 Micropatterning, collagen overlay, and live cell microscopy results

Portions of the patterns that had a majority of MDA-MB-231 cells on the patterns we compared to the regions containing one or more fibroblasts (Fig 13).

Figure 13: Fibroblasts (red) pattered with MDA-MB-23. Scale bar represents 50 μm (20x).

There were instances where MDA-MB-231 cells (green) polarized and aligned with fibroblasts when cultured together in a collagen gel (Fig 14). The collagen gel

33 surrounding the fibroblasts seemed to be extra dense and collagen fibers appeared aligned according to the fibroblasts in comparison to the parts of the gel without fibroblasts (Fig.

14 & 14L).

Figure 14: Cellular interactions in gels between fibroblasts (green) and MDA-MB-231(red). (L)The bright field image showed collagen gel contraction near the fibroblasts. (R)The corresponding fluorescent image. (20x) Scale bar is 50μm.

Cancer cells were observed to extend filapodia towards another cell within the collagen matrix and move towards it (Fig. 15). Initially, the cancer cell had several smaller cytoskeletal projections, but the after 20 mins, the projections merged into one cell process. The cancer cell adopted more of a spindle shape form as it was moving towards the other cell. After ten minutes, the cancer cell retracted its cytoskeletal project and changed direction, moving towards other cells. A blob-like extension formed from the reverse side as the cancer cell traveled towards the other cells.

34

t = 2h t = 3:30 h t = 4:30 h

t = 7:15 h t = 5:30 h t = 9 h Figure 15: Cancer cell migration and extension of cytoskeletal projections was observed over a period of 7 hours. (10x) Scale bar is 100μm.

Unlike the overt cell migration observed in the collagen gel, cells adhered to the fibronectin patterns were not observed to migrate along the 2D surface. The fibroblasts and tumor cells that attached to patterns moved within the circle pattern, but did not migrate out of the pattern. Over time, the cells were observed to spread on the circle patterns, making them difficult to visualize (Fig. 16).

35

t = 3 h t = 6 h

t = 3 h t = 12 h t = 12 h

Figure 16: Live cell microscopy with fibroblasts (red) and cancer cells on PDMS substrate showing that cells do not interact via migration without the presence of collagen gel. (20x) Scale bar is 100μm.

After the initial trial applying the PLL-glutaraldehyde treated gasket, substantial cell death was observed. It was theorized that the factor contributing to cell toxicity was glutaraldehyde. Therefore, the dilute collagen solution was added to the gasket an allowed to incubate for 24 hours. The collagen solution leached excess glutaraldehyde from the gasket and lowered cytotoxicity. In comparison to the previous trials with glutaraldehyde treated gaskets to the trials with the precoating of collagen solution, reduced cell toxicity was observed.

A MATLAB code was successfully written to calculate the distance between the center of each fibroblast and to each tumor cell. Images of the MATLAB process are

36 shown below starting with the fluorescent original image (Fig. 17a) being split into two channels, the cancer cells (Fig. 17b) and fibroblasts (Fig. 17c), respectively. Then the histogram of the distances was plotted comparing the time right after the addition of the collagen gel overlay to the end of the experiment, at time = 36 hours (Fig. 18). No significance was found using a two-tailed Student t-test assuming equal variance

(p>0.05). An a priori power analysis test was run and the total sample size necessary to achieve adequate power with an α error probability at 5% and 1-β error probability at

95% was determined to be 42. A frequency distribution of a 7x7 grid of circle patterns taken at 20x was created as a model distribution for patterned cells (Fig. 19). Distances were calculated from the central pattern to each of the other patterns. Cancer cells were also culture with the absence of fibroblasts. There was no significance found between movement of cancer cells (Fig. 20).

a b

c

Figure 17: (a) The original image with patterned fibroblasts (green) and cancer cells (red) was split into (b) a cancer cell channel and (c) fibroblast channel. (20x) Scale bar is 100μm.

37

250

200

150 Time = 1h

100 Frequency 50

0 0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 Distance (m)

250

200

150 Time = 36h

100 Frequency 50

0 0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 Distance (m)

Figure 18: Histograms show the calculated distances from each fibroblast to each respective cancer cell from the initial positions to the end of the experiment at 36h after adding the collagen gel. (N=14 for each condition).

10

8

6

4 Frequency 2

0

150 200 250 300 350 400 450 500 550 600 650 700 750 800 Distance (m)

Figure 19: A histogram of the theoretical calculated distances from one pattern to every other pattern for a 7x7 grid of circle patterns taken at 20x.

38

20 Time = 1h Time = 36h 15

10 Frequency 5

0 0 0 50100150200250300350400450500550 50100150200250300350400450500550 Distance (m)

Figure 20: A histogram of cancer cells cultured without fibroblasts. No significance was found (p<0.05).

In addition to calculating the histogram of all of the cells, individual cells were also isolated. The distance from one fibroblast to each tumor cell was compared over time

(Fig 21). In some cases, there was significance found in the change in distance, where the distances from the tumor cells to the fibroblasts decreased.

39

4 Time = 1h Time = 36h 3

2 Frequency 1

0

80 80 100 120 140 160 180 200 100 120 140 160 180 200 Distance (m)

4 Time = 1h Time = 36h 3

2 Frequency 1

0

50 50 100 150 200 250 300 350 100 150 200 250 300 350 Distance (m)

Figure 21: Histogram of the distances from an individual fibroblast to each tumor cell in the field of view over time. (L) There was found to be a significant change in the tumor cell distance compared to the fibroblasts (p < 0.05). (R)There was no significant change in fibroblast and tumor cell distance (p>0.0).

40

5.4 Myofibroblast differentiation

The condition that resulted in the most myofibroblast differentiation was found to be 10ng/mL in 1% serum DMEM replaced every two days for five days gave an optimal myofibroblast differentiation (Fig. 22).

Figure 22: Myofibroblasts (orange-yellow) and fibroblasts (green) stained with Phalloidin, DAPI, and α- SMA.Scale bar is 100μm (20x).

The myofibroblast differentiation images were successfully thresholded and denoised (Fig 23). The individual cells were isolated via the blob segmentation method and counted. The percentage of the total number of cells expressing α-SMA was calculated for the conditions for serum starvation for 24h, followed by addition of 5, 10, or 20ng/mL TGF-β in DMEM with 1% serum changed every 2 days (Fig 23). A one-way

ANOVA test followed by a Tukey-Kramer test showed that 10 ng/mL of TGF-β was found to be significant between the rest of the conditions (p<0.05).

41

a b

c d

Figure 23: MATLAB images analysis of the percentage of myofibroblast differentiation. a) The original channel overlay with phalloidin 488 (green), DAPI (blue), and yellow-orange staining for the α -SMA. b) The thresholded, denoised image of the α-SMA channel c) isolation of each cell and d) counting the total number of cells by the nuclei.

Figure 20: The percentage of fibroblasts that expressed A one-way Anova was performed followed by a Tukey’s multiple comparison test (p<0.05).

42

CHAPTER 6: DISCUSSION

6.1 Cell seeded collagen gels within a collagen matrix

The first method, based on Provenzano’s method, involving two cell seeded collagen gels enveloped within an acellular collagen matrix was not deemed optimal for this system because it did not meet the design criterion of spatial constraint. Specifically, more control over initial cell placement was desired between the tumor and fibroblasts.

This CSCG system provided information about a population of cells in a matrix, but a study of single cells at specified distances was preferred. By focusing on individual cells, image analysis on the cells could be simplified as well as having clear observations of individual cell morphologies. Having a mass effect of cells would substantially increase the difficultly to identify individual cell interactions. Therefore, due to these constraints, a new approach to allow for single cell study was pursued.

6.2 Micropattern cell seeding optimization

In order to fulfill the design criteria of optimizing stamped homogenous fibronectin islands and reducing non-specific cell binding to the substrate, various factors had to be tested such as the type of substrate, initial cell seeding density, fibroblast to tumor cell seeding ratio, and seeding time. Initially, two types of substrates were tested: glass and tissue culture treated plastic. Fibronectin failed to adhere well to the glass substrate due to its high hydrophobicity that inhibited protein adsorption, and resulted in no observable patterns. In comparison, the hydrophilicity of tissue culture plastic allowed for slight fibronectin adsorption, but still did not produce repeatable, homogenous patterns.

43

Since the tissue culture plastic and glass substrate options were not viable, a

PDMS surface was explored. PDMS is an inert, silicon-based organic polymer that has linear, viscoelastic, and isotropic properties. Because of the hydrophobic nature of the

PDMS limits its use with biological specimens, it needs to be surface modified to be either less hydrophobic or more hydrophilic in order to allow the fibronectin patterns to bind to the surface. One straightforward method to reduce the hydrophobicity of PDMS is to use oxygen plasma. Specifically, treating PDMS with oxygen plasma reduces its hydrophobicity to allow protein adsorption. The mechanism of oxygen-plasma treatment works by applying the reactive species, atomic oxygen, that attacks the PDMS’s siloxane backbone to form polar silanol Si-OH surface bonds and an oxygen rich SiOx silica-like layer [48]. The polar silanol (SiOH) groups replace the -O-Si(CH3)2- and effectively convert the hydrophobic surface to slightly hydrophilic, thus becoming a more viable substrate for protein adsorption. A disadvantage of oxygen plasma method is that the treated PDMS will recover its original hydrophobicity over time. Therefore, fibronectin was stamped immediately onto the treated surface. Oxygen-plasma treated PDMS proved to be a viable substrate that resulted in repeatable, homogenous fibronectin patterns.

The discrepancies between the quality of fibronectin patterns on the three substrate types was governed by the principles of protein adsorption. Intermolecular forces, substrate hydrophobicity, and electrostatic interactions are the principal forces that drive protein adsorption, or the accumulation of proteins on a surface. Specifically, proteins have more affinity to hydrophobic surfaces rather than hydrophilic surfaces.

Glass is neutrally charged and hydrophobic, while standard tissue culture treated polystyrene plastic is negatively charged and hydrophilic [49]. The combination of the

44 tissue culture polystyrene’s negative charge and hydrophilicity cause its affinity for protein adsorption to be low. Since like charges repel each other, the negatively charged tissue culture polystyrene will repel the negatively charged fibronectin and will not allow for optimal fibronectin patterns. Furthermore, the neutrally charged glass substrate did not express affinity for the fibronectin to adsorb. The hydrophobic and electrostatic effects of the polystyrene and glass substrates contributed to the poor adsorption of fibronectin and unsatisfactory patterns. Therefore, by oxygen-plasma treating the PDMS and making it slightly hydrophilic, fibronectin was attracted to the substrate and resulted in optimal patterns. Since oxygen-plasma treated PDMS resulted in the highest quality patterns, further experiments used this substrate.

After the repeatable patterns were achieved, the optimal cell seeding density was determined to be 100,000 cells in each well of a six well plate. Since probability played a major factor in the tumor cells and fibroblasts landing and attaching to a fibronectin pattern, the initial seeding density, fibroblast-to-tumor cell ratio, as well as the seeding time were important considerations. In order to reduce the chances of more than one cell binding to the same pattern, these variables were tested. One cell per pattern was optimal in order to observe the interactions between single cells. As the number of initial cells seeded increased, there was a higher ratio between cells and patterns. With more cells than patterns, the probability of more than one cell attaching to one pattern increased. The ratio of seeded fibroblasts to tumor cells was also important because more tumor cells than fibroblasts were desired because specific interest was taken in the effect of one fibroblast on surrounding tumor cells. Seeding more tumor cells than fibroblasts increased the probability that more tumor cells would attach to patterns than fibroblasts.

45

The ratio of fibroblasts and tumor cells seeded affected the ratio of fibroblasts to tumor cells on the patterns. With a greater ratio, there would be too few fibroblasts to many tumor cells; with a lesser ratio, there would be almost an equal number of fibroblasts to tumor cells. Therefore, a ratio of one fibroblast to three tumor cells was chosen.

In addition to seeding time and density, the length of time cells were left to attach to patterns before washing away excess cells also affected the number of patterns had one or more cells. The time it takes for cells to adhere to the substrate depends on membrane adhesion ligand-initiating signaling cascades involving integrin receptors and formation of focal adhesion complexes. Furthermore, the attraction of cells to the substrate via electrostatic interactions, Van der Waals forces, and protein-surface distances also affect the adhesion time [10]. Ligand size and shape influence the competition between adhesion proteins, and thus play a part in the regulation of cell membrane adhesion to the substrate [11]. Since human dermal fibroblasts are larger than the MDA-MB-231 adenocarcinoma cells, the greater area allows for greater chance of membrane bound adhesion ligands to attach to a substrate. Therefore, this size difference relating to exposure of adhesion ligands could be a reason as to why dermal fibroblasts were observed to adhere faster than the MDA-MB-231 cells on all the substrates tested. The interplay between the seeding density and the seeding time were the critical factors in increasing the chances of one cell attaching to one pattern. Some reasons why more than one cell was found on patterns were that the combination of the seeding density and seeding time increased the probability for cells to attach to the same pattern. Another reason for one pattern having more than one cell is that cells could undergo mitosis and

46 proliferate on the patterns. Cell proliferation was one reason why the duration of the assay was kept to only 1.5 to 2 days.

An additional issue that was observed was the prevalence of nonspecific cell binding to the substrate areas between patterns. Due to both fibroblast and tumor cell affinity to tissue culture plastic and the fibronectin patterns, cells would adhere to both the plastic and the patterns. This lack of clearly visible patterned cells essentially nullified the whole reason for patterns providing cells with initial positions. Therefore, a blocking agent to prevent cell adhesion to non-fibronectin islands was desired. Pluronic acid F127 was used to surface modify the PDMS to block protein adsorption. Pluronics® is a part of a group of poly(ethylene glycol) and poly(ethylene oxide)/poly(propylene oxide)/poly(ethylene oxide) (PEO-PPO-PEO) tri-block copolymers that inhibit protein adsorption [50]. The amphipathic copolymers of Pluronics®, in addition to its lack of ionic charge, and high hydrophilicity contribute to its protein adsorption inhibitory properties [51]. Specifically, the PPO chain will attach to hydrophobic surfaces through hydrophobic interactions, while the hydrophilic PEO chains extend into the medium [52].

Through blocking with Pluronics®, cells only attached to the fibronectin and not to the areas with Pluronics®.

One challenge was keeping the collagen gel attached to the Pluronics® cured

PDMS. Maintaining contact between the collagen gel and the cells was critical to this system in order to allow for the cells at spatially constrained distances to interact with the

3D collagen matrix. The lack of collagen gel adhesion to the PDMS surface was probably overlay, the cells would not be able to interact with the 3D in vivo mimic of a collagen matrix, thus not fulfilling the design criteria of this system due to two reasons: 1) the high

47 hydrophobicity of the PDMS prohibiting protein adsorption and 2) the application of

Pluronics® f127.

Initially, the lack of collagen adhesion to the PDMS substrate was thought to be due to Pluronics® f127 being applied to the whole PDMS substrate. Therefore,

Pluronics® f127 was only applied on the stamped portion of the PDMS with the hope that the collagen would attach to the rest of the PDMS substrate uncoated with

Pluronics® f127. However, the collagen gel was still unable to securely maintain attachment to the substrate, most likely due to the hydrophobicity of the PDMS. With this challenge of holding the collagen gel in place over the cells, another method of binding the collagen gel to lay on top of the patterned cells was attempted.

A PDMS gasket was fabricated and placed around the patterned area, where it covalently interacts with the PDMS substrate and formed a seal. This gasket was treated with poly-L-lysine (PLL) and glutaraldehyde, which would crosslink with the collagen matrix and physically hold it in place. Specifically, PLL is a homopolymer of the amino acid L-lysine and contains a positively charged amine group which crosslinks with glutaraldehyde. Glutaraldehyde is an organic amine-reactive crosslinker typically used to fix proteins and peptides, and is also cytotoxic to cells. The PLL/glutaraldehyde treatment successfully bound the collagen gel to the gasket and did not allow the gel to float. This way, the collagen gel was kept in close proximity to the cells and allowed them to interact with a 3D environment.

Substantial cell death was observed after application of the PLL-glutaraldehyde treated gasket, as compared reduce number of dead cells after coating the gasket with a dilute collagen solution. This difference in healthy cell count was attributed to exposure

48 to the cytotoxic agent, glutaraldehyde. In order to reduce the number of dead cells, a method to reduce the cytotoxicity of glutaraldehyde by adding a dilute collagen layer to leach out excess glutaraldehyde was attempted. By only applying glutaraldehyde to the inside walls of the gasket, instead of allowing it to be absorbed around the entire surface, the amount of glutaraldehyde introduced into the system was reduced. The rationale behind this method is that since some glutaraldehyde leached into the PDMS gasket during incubation, washing repeatedly with water did not remove the absorbed glutaraldehyde. This absorbed glutaraldehyde would leach out after gasket application to the system and cause cell death. Therefore, soaking the gasket in a dilute solution of collagen for at least a day would leach out and react with residual glutaraldehyde, thus reducing the levels of cytotoxic glutaraldehyde that will come in contact with cells. No significant difference between the total number distances between the cells at the initial time and when the experiment was ended 36 hours after was found. This could be due to there not being enough time for the cells to interact or the concentration of collagen was too much and hindered cell movement. However, when individual fibroblasts were analyzed compared to the tumor cells, there was significance found in some cases. This could be due to the forces exerted by each fibroblast causing a tumor cell to migrate towards it, and away from another fibroblast, thus skewing the histogram displaying the total distances.

6.4 Myofibroblast differentiation

Fibroblast differentiation into myofibroblasts is TGF-β concentration dependent and requires cell adhesion-dependent signals and Smad2 phosphorylation [53]. Studies have shown that TGF-β initiates the gene transcription of α-SMA and results in an

49 increased expression of α-SMA in fibroblasts [54]. These differentiated fibroblasts have increased production of extracellular matrix components, mainly of collagen types I, III, and IV. The signaling pathway that TGF- β activates is the Smad pathway, where TGF-

β-dependent cell proliferation is dependent on the presence of Smad2 and Smad 3.

Specifically, Smad 2 and Smad3 are central in the TGF- β signaling that results in the activation of specific genes relating to α-SMA and the accumulation of Smad complexes in the nuclear region [55]. Therefore, the TGF- β concentration necessary to activate the

Smad pathway to stimulate the production of α-SMA proteins had to be determined in order to differentiate myofibroblasts from fibroblasts. If TGF- β concentrations are too low, there might not be enough to bind to RI and RII receptors in order to stimulate the cascade of the Smad signaling pathway, impeding the differentiation of fibroblasts.

Many studies used 10ng/mL as the optimal concentration of TGF- β for fibroblast differentiation, which matched the concentration that resulted in the most myofibroblasts found in this thesis [56] [57]. Even though the experimental percentage of cells expressing α-SMA did not meet the originally desired percentage of 60%, it is unknown if the 10% difference in experimental and desired percentage of differentiated cells will have significant impact on the results. One reason that the experimental percentage was lower than the desired percentage of α-SMA expressing cells could be due to the combined effects of the time length of incubation with TGF-ß and TGF-ß concentration with this specific cell type.

50

6.5 Design constraints

Several design constraints were presented in this system such as cell attachment, reduction of non-specific cell binding, cell death, and availability of lab equipment.

Optimizing cell attachment to the fibronectin patterns as well as reducing non-specific cell adhesion to the substrate was troubleshot. By varying the cell attachment time, ratio of fibroblasts to cancer cells, type of substrate, cell seeding densities, and surface treating the substrate, favorable cell attachment to the fibronectin patterns with simultaneous decrease in non-specific cell adhesion was observed. By varying the prior parameters, and optimizing the total number of fibronectin patterns that were filled with cells to at least 70%, the design goal of spatial constraint of cells was achieved. Furthermore, cell death was an omnipresent issue that was caused by various factors. Specifically, one cause could be due to over-drying the cells and substrate before application of the PDMS gasket. The substrate was allowed to air dry because the PDMS gasket was observed to adhere to the dry PDMS substrate. When the PDMS substrate was slightly damp, the gasket did not stay firmly in one place, but rather slid horizontal along the substrate as well as floating upwards after fresh media was added. By minimizing the amount of drying time, cells were not allowed to dry out and die. Furthermore, the glutaraldehyde treatment proved to be cytotoxic for the cells. Therefore, excess glutaraldehyde was leached out for 24 hours in a dilute collagen solution, as mentioned above. Another design constraint was the availability of equipment, seeing that the most accessible microscope was the Leica CTR 6500 fluorescent microscope. The system was designed to be compatible with the CTR 6500 for data acquisition through bright field, fluorescence, and live cell microscopy. A final constraint was the ability of the image

51 processing algorithms to successfully perform blob segmentation on the images and separate and identify each cell, while excluding noise. The Otsu algorithm uses an automatic threshold that minimizes the weighted intra-class variance, while maximizing the inter-class variance, to give the optimal contrast. The foreground and background intensity are calculated and clustered into a bimodal histogram, then calculates the optimal threshold to separate the foreground and background based on the variance. In some cases, even with the application of the Otsu threshold in combination with denoising techniques, not all of the cells are accounted for or noise is counted as a cell.

Therefore, there are implicit errors due to the current state of image processing, which could skew the calculated distances between cells.

52

CHAPTER 7: SIGNIFICANCE & FUTURE DIRECTIONS

The goal to develop a customizable 3D system that mimics native tissue that would allow for the study of the individual intercellular interactions of tumor and fibroblasts was successful. The system was validated and the device goals and design criteria were largely met. Specifically, the criterion of spatially constraining at least 70% of cells at an initial starting position was succeeded through the technique of micropatterning. Next, the 3D aspect of the in vivo microenvironment was simulated by the introduction of a 2.0 mg/mL collagen gel overlaid to allow for cell-cell and cell- matrix interations and invasion. Cellular interaction data was gathered in the form of tracking cell movement over time and processing this via MATLAB, showing that fibroblasts do effect tumor cell movement in a matrix. Finally, the optimization of differentiation myofibroblasts for the future inclusion into this system was done. Even though the quota of having at least 60% of fibroblasts expressing α-SMA was not achieved, it is unknown if this discrepancy will have a significant biological impact when included into the system. Studying cells on the individual level instead of in groups is important to understand their activities on a basic level, which then could be the stepping stone to understanding group behaviors of cells. This system is significant for its contribution to the gradual movement away from studying cells in traditional, simplistic

2D cell culture systems to 3D systems that better mimic the in vivo microenvironment, as well as offering a technique to spatially constrain cells to allow for control and tracking of cell interactions over time.

Even though this system offers great capability in studying interactions between tumor and fibroblasts over time at spatially constrained initial positions, there are some

53 potential drawbacks and pitfalls. For example, this system does not fully emulate the in vivo 3D tumor microenvironment, which could result in differences between observed cell behavior in the system and actual cell behavior in the body. For future experiments, the PDMS substrate could be substituted with a softer material such as a polyacrylamide gel to better mimic the stiffness of the in vivo environment. Even though the Young’s modulus (E) of PDMS softer than glass or plastic substrates, it is still greater than that of native tissue. Native tissue is between 360-870 kPa while for a PDMS base to crosslinker ratio of 1:10, the elastic modulus is 580 kPa [58]. Another drawback of this device is that even with the application of the PDMS gasket to hold the collagen gel in place, the collagen overlay is not directly attached to the PDMS substrate—it is just physically being held down. For future system development, a method to tightly attach the collagen gel to the substrate in order to prevent any accidental detachment should be researched.

This 3D micropatterned system allows for much customization, allowing for various parameters to be changed for future experiments. In particular, parameters such as cell type and style of micropatterns can be altered. Myofibroblasts from the optimized fibroblast differentiation protocol can be included into the system to study the effects of introducing a more contractile cell type that are found in the tumor microenvironment as cancer-associated fibroblasts. Also, the effects of a primary cell line such as primary human mammary fibroblasts can be studied. Another aspect that can be changed for future studies is type of micropatterns used. In this thesis, circle patterns were used for system development, however, it would be interesting to use different sizes and geometric shapes such as squares, rectangles, and crossbows to study the effect geometries on cell behavior. Furthermore, the orientations and distances between each

54 micropattern can be altered and how distance between cells affects intercellular interactions can be observed. Finally, contractility inhibitors such as the ROCK inhibitor

Y-27632 or the MLCK inhibitors ML-7 and/or ML-9 can be added. Since fibroblasts and myofibroblasts are particularly contractile cells, rendering them unable to fully contract is will affect the extent of how they remodel the ECM and interact with tumor cells.

Therefore, a future study can compare the differences between contractility inhibited and uninhibited fibroblasts and myofibroblasts in the system.

The implications of designing a versatile 3D system with a collagen overlay will allow for the improved study of cells in a microenvironment that more closely resembles the native tissue, but also offers control over spatial conditions of cells. This system can be adapted to a wide range of practical usages in almost biological event that involve cells interaction with each other. For example, not only can tumor cell migration be observed by studying the interactions between fibroblasts and tumor cells, but a possible application of this device is to also study wound contraction in granulation tissue between fibroblasts and epithelial cells. The future of modeling biological processes involving cellular interactions lie with the systems that better mimic the in vivo environment.

Therefore, this study stands at the intermediary point between traditional 2D cell culture and in vivo studies by offering not only a system design, but also a range of technique and tools to assist in the study of intercellular interactions in a tissue analog.

55

LIST OF REFERENCES

1. Allinen, M., et al., Molecular characterization of the tumor microenvironment in breast cancer. Cancer Cell, 2004. 6(1): p. 17-32.

2. Geckil, H., et al., Engineering hydrogels as extracellular matrix mimics. Nanomedicine (Lond), 2010. 5(3): p. 469-84.

3. Jeon, H., E. Kim, and C.P. Grigoropoulos, Measurement of contractile forces generated by individual fibroblasts on self-standing fiber scaffolds. Biomed Microdevices, 2011. 13(1): p. 107-15.

4. Chambers, A.F., A.C. Groom, and I.C. MacDonald, Dissemination and growth of cancer cells in metastatic sites. Nat Rev Cancer, 2002. 2(8): p. 563-72.

5. Nyberg, P., T. Salo, and R. Kalluri, Tumor microenvironment and angiogenesis. Frontiers in bioscience: a journal and virtual library, 2007. 13: p. 6537-6553.

6. Paszek, M.J., et al., Tensional and the malignant phenotype. Cancer Cell, 2005. 8(3): p. 241-54.

7. Hanahan, D. and J. Folkman, Patterns and emerging mechanisms of the angiogenic switch during tumorigenesis. cell, 1996. 86(3): p. 353-364.

8. Butler, T.P., F.H. Grantham, and P.M. Gullino, Bulk transfer of fluid in the interstitial compartment of mammary tumors. Cancer Res, 1975. 35(11 Pt 1): p. 3084-8.

9. Dvorak, H.F., Tumors: wounds that do not heal. Similarities between tumor stroma generation and wound healing. N Engl J Med, 1986. 315(26): p. 1650-9.

10. Kalluri, R. and M. Zeisberg, Fibroblasts in cancer. Nat Rev Cancer, 2006. 6(5): p. 392-401.

11. Shieh, A.C., et al., Tumor cell invasion is promoted by interstitial flow-induced matrix priming by stromal fibroblasts. Cancer Res, 2011. 71(3): p. 790-800.

12. Lamouille, S. and R. Derynck, Cell size and invasion in TGF-beta-induced epithelial to mesenchymal transition is regulated by activation of the mTOR pathway. J Cell Biol, 2007. 178(3): p. 437-51.

13. Stacker, S.A., et al., VEGF-D promotes the metastatic spread of tumor cells via the lymphatics. Nat Med, 2001. 7(2): p. 186-91.

14. Kessenbrock, K., V. Plaks, and Z. Werb, Matrix metalloproteinases: regulators of the tumor microenvironment. Cell, 2010. 141(1): p. 52-67.

56

15. Poincloux, R., et al., Contractility of the cell rear drives invasion of breast tumor cells in 3D Matrigel. Proc Natl Acad Sci U S A, 2011. 108(5): p. 1943-8.

16. Sugiura, H., et al., 3-Nitrotyrosine inhibits fibroblast-mediated collagen gel contraction and chemotaxis. Eur Respir J, 2009. 34(6): p. 1452-60.

17. Wolf, K., et al., Multi-step pericellular proteolysis controls the transition from individual to collective cancer cell invasion. Nat Cell Biol, 2007. 9(8): p. 893- 904.

18. Provenzano, P.P., et al., Contact guidance mediated three-dimensional cell migration is regulated by Rho/ROCK-dependent matrix reorganization. Biophys J, 2008. 95(11): p. 5374-84.

19. Winer, J.P., S. Oake, and P.A. Janmey, Non-linear elasticity of extracellular matrices enables contractile cells to communicate local position and orientation. PLoS One, 2009. 4(7): p. e6382.

20. Byfield, F.J., et al., Endothelial actin and cell stiffness is modulated by substrate stiffness in 2D and 3D. J Biomech, 2009. 42(8): p. 1114-9.

21. Cukierman, E., R. Pankov, and K.M. Yamada, Cell interactions with three- dimensional matrices. Curr Opin Cell Biol, 2002. 14(5): p. 633-9.

22. Bao, G. and S. Suresh, Cell and molecular mechanics of biological materials. Nat Mater, 2003. 2(11): p. 715-25.

23. Chen, C.S., et al., Micropatterned surfaces for control of cell shape, position, and function. Biotechnol Prog, 1998. 14(3): p. 356-63.

24. Discher, D.E., P. Janmey, and Y.L. Wang, Tissue cells feel and respond to the stiffness of their substrate. Science, 2005. 310(5751): p. 1139-43.

25. Pelham, R.J., Jr. and Y. Wang, Cell locomotion and focal adhesions are regulated by substrate flexibility. Proc Natl Acad Sci U S A, 1997. 94(25): p. 13661-5.

26. Yeung, T., et al., Effects of substrate stiffness on cell morphology, cytoskeletal structure, and adhesion. Cell Motil Cytoskeleton, 2005. 60(1): p. 24-34.

27. Tomasek, J.J. and E.D. Hay, Analysis of the role of microfilaments and microtubules in acquisition of bipolarity and elongation of fibroblasts in hydrated collagen gels. J Cell Biol, 1984. 99(2): p. 536-49.

28. Friedl, P., Prespecification and plasticity: shifting mechanisms of cell migration. Curr Opin Cell Biol, 2004. 16(1): p. 14-23.

57

29. Zaman, M.H., et al., Migration of tumor cells in 3D matrices is governed by matrix stiffness along with cell-matrix adhesion and proteolysis. Proc Natl Acad Sci U S A, 2006. 103(29): p. 10889-94.

30. Lo, C.M., et al., Cell movement is guided by the rigidity of the substrate. Biophys J, 2000. 79(1): p. 144-52.

31. Gray, D.S., J. Tien, and C.S. Chen, Repositioning of cells by mechanotaxis on surfaces with micropatterned Young's modulus. J Biomed Mater Res A, 2003. 66(3): p. 605-14.

32. Brodsky, B. and E.F. Eikenberry, Characterization of fibrous forms of collagen. Methods Enzymol, 1982. 82 Pt A: p. 127-74.

33. Provenzano, P.P., et al., Collagen density promotes mammary tumor initiation and progression. BMC Med, 2008. 6: p. 11.

34. Provenzano, P.P., et al., Collagen reorganization at the tumor-stromal interface facilitates local invasion. BMC Med, 2006. 4(1): p. 38.

35. Santimaria, M., et al., Immunoscintigraphic detection of the ED-B domain of fibronectin, a marker of angiogenesis, in patients with cancer. Clin Cancer Res, 2003. 9(2): p. 571-9.

36. Postlethwaite, A.E., et al., Induction of fibroblast chemotaxis by fibronectin. Localization of the chemotactic region to a 140,000-molecular weight non- gelatin-binding fragment. J Exp Med, 1981. 153(2): p. 494-9.

37. Kimlin, L.C., G. Casagrande, and V.M. Virador, In vitro three-dimensional (3D) models in cancer research: an update. Mol Carcinog, 2013. 52(3): p. 167-82.

38. Yu, H., J.K. Mouw, and V.M. Weaver, Forcing form and function: biomechanical regulation of tumor evolution. Trends Cell Biol, 2011. 21(1): p. 47-56.

39. Wang, N., et al., Micropatterning tractional forces in living cells. Cell Motil Cytoskeleton, 2002. 52(2): p. 97-106.

40. Gefen, A. and B. Dilmoney, Mechanics of the normal woman's breast. Technol Health Care, 2007. 15(4): p. 259-71.

41. Roeder, B.A., et al., Tensile mechanical properties of three-dimensional type I collagen extracellular matrices with varied microstructure. J Biomech Eng, 2002. 124(2): p. 214-22.

42. Lopez-Garcia, M.D., D.J. Beebe, and W.C. Crone, Young's modulus of collagen at slow displacement rates. Biomed Mater Eng, 2010. 20(6): p. 361-9.

58

43. Arora, P.D., N. Narani, and C.A. McCulloch, The compliance of collagen gels regulates transforming growth factor-beta induction of alpha-smooth muscle actin in fibroblasts. Am J Pathol, 1999. 154(3): p. 871-82.

44. Stuelten, C.H., et al., Transient tumor-fibroblast interactions increase tumor cell malignancy by a TGF-Beta mediated mechanism in a mouse xenograft model of breast cancer. PLoS One, 2010. 5(3): p. e9832.

45. Grotendorst, G.R., H. Rahmanie, and M.R. Duncan, Combinatorial signaling pathways determine fibroblast proliferation and myofibroblast differentiation. Faseb j, 2004. 18(3): p. 469-79.

46. Vaughan, M.B., E.W. Howard, and J.J. Tomasek, Transforming growth factor- beta1 promotes the morphological and functional differentiation of the myofibroblast. Exp Cell Res, 2000. 257(1): p. 180-9.

47. Masur, S.K., et al., Myofibroblasts differentiate from fibroblasts when plated at low density. Proc Natl Acad Sci U S A, 1996. 93(9): p. 4219-23.

48. Tan, S.H., et al., Oxygen plasma treatment for reducing hydrophobicity of a sealed polydimethylsiloxane microchannel. Biomicrofluidics, 2010. 4(3): p. 32204.

49. Ramsey, W.S., et al., Surface treatments and cell attachment. In Vitro, 1984. 20(10): p. 802-8.

50. Amiji, M. and K. Park, Prevention of protein adsorption and platelet adhesion on surfaces by PEO/PPO/PEO triblock copolymers. Biomaterials, 1992. 13(10): p. 682-92.

51. Bailey, F. and J. Koleske, Configuration and hydrodynamic properties of the polyoxyethylene chain in solution. 1966, Marcel Dekker, Inc New York. p. 794-5.

52. Kayes, J.R., DA., Adsorption characteristics of certain polyoxyethylene- polyoxypropylene block co-polymers on polystyrene latex. Colloid & Polymer Sci, 1979(257): p. 622-629.

53. Thannickal, V.J., et al., Myofibroblast differentiation by transforming growth factor-beta1 is dependent on cell adhesion and integrin signaling via focal adhesion kinase. J Biol Chem, 2003. 278(14): p. 12384-9.

54. Roberts, A.B., et al., Transforming growth factor-beta: possible roles in carcinogenesis. Br J Cancer, 1988. 57(6): p. 594-600.

55. Evans, R.A., et al., TGF-beta1-mediated fibroblast-myofibroblast terminal differentiation-the role of Smad proteins. Exp Cell Res, 2003. 282(2): p. 90-100.

59

56. Shi, Y., et al., Substrate stiffness influences TGF-beta1-induced differentiation of bronchial fibroblasts into myofibroblasts in airway remodeling. Mol Med Rep, 2013. 7(2): p. 419-24.

57. Hu, Y.L., et al., [The effect of RhoA/Rho kinase signal pathway on TGF-beta1- induced phenotypic differentiation of human dermal fibroblasts]. Zhonghua Zheng Xing Wai Ke Za Zhi, 2011. 27(5): p. 376-80.

58. Park, J.Y., Yoo, S.Y., Lee, E., Lee, D.H., Kim, J.Y., Lee, S., Increased poly(dimethylsiloxane) stiffness improves viability and morphology of mouse fibroblast cells. BiocChip Journal, 2010. 4(3): p. 230-236.

60

APPENDIX A

MATLAB code to find the distances between tumor and cancer cells and plot the results as a histogram

%find distance between red and green cells clc,clear,close all img = imread('12 22 2012 1 to 10 6 hours_20x tile2.tif'); img2= imread('1_7 Circle Feb 10 2013 2nd exp 20X.tif'); figure; imshow(img); figure; imshow(img2); %% img= imread('12 22 2012 1 to 10 6 hours_20x tile2.tif'); figure; imshow(img); %im1 = img1(:,:,1); %cancer cells (RED) im2 = img1(:,:,2); %fibroblasts (GREEN) im1 = img(:,:,1); %cancer cells im2 = img(:,:,2); %fibroblasts %imshow(img); im1 = img(:,:,1); %cancer cells figure; imshow(im1); im2 = img(:,:,2); %fibroblasts figure; imshow(im2); level1 = graythresh(im1); level2 = graythresh(im2); new1 = im2bw(im1,level1); new2 = im2bw(im2,level2); imshow(new1); figure; imshow(new2);

%if there's too much white spot noise, then use a new threshold if length(find(new2==1))>50000 new2 = im2bw(im2,level1); end

%label the fibroblasts & cancer cells separately [CC N]=bwlabel(new1); [Fib num]=bwlabel(new2); for iR = 1:N % filter cancer cell image Area = length(find(CC==iR)); if Area <= 20 CC(CC==iR)=0; end end [CC N]=bwlabel(CC); for iF = 1:num % filter fibroblast image Area = length(find(Fib==iF)); if Area <= 200 Fib(Fib==iF)=0; end end [Fib num]=bwlabel(Fib); stat_Fib = regionprops(Fib,'Centroid');%centroids of fibroblast stat_CC = regionprops(CC,'Centroid');%centroids of cancer cells

61

D = zeros(N,num); %row = dist btwn designated fibroblast and cancer cells; %column = different fibroblasts for j = 1:num %picks the fibroblast one at a time for i = 1:N %calculates the distance from each fibroblast to each cancer cell cR = stat_CC(i).Centroid; cF = stat_Fib(j).Centroid; dx = cF(1)-cR(1); dy = cF(2)-cR(2); dist = sqrt(dx^2+dy^2); %cartesian distance formula D(i,j)=dist; end end distance_um= D./2.08; %Convert the pixels into um figure; hist(distance_um(:,1)); figure; hist(distance_um(1,:)); figure; hist(distance_um); %title('Histogram of the distances between each fibroblasts to each cancer cell'); xlabel('Distance (um)'); ylabel('Frequency');

MATLAB code to find the percentage of cells expressing α-SMA function out = SMA_analysis clc, clear, close all sdir = 'SMA'; % folder name tif = dir([sdir '/*.tif']);%file names in a list Count = [];%empty matrix for i = 1:length(tif) Image = [sdir '/' tif(i).name];%one file i.e. image at a time NUM_SMA = SMA_count(Image); %count SMA-expressing cells per image NUM_DAPI = DAPI_count(Image);%count all of cells per image per_SMA = NUM_SMA/NUM_DAPI*100;%calculate percent of SMA cells TOT = [NUM_SMA NUM_DAPI per_SMA];%add data to list Count = [Count;TOT]; end out = Count; function output = SMA_count(image) img = imread(image); imgn = imresize(img(:,:,1),0.3); %figure(2);imshow(imgn); %level = graythresh(imgn); im = im2bw(imgn,.1); im2 = imclearborder(im); %clear the cells around the border %figure(3); imshow(im2); [im3 N]= bwlabel(im2,4);

62

%Remove noise for i = 1:N Area = length(find(im3 == i)); if Area < 800 && Area > 1000 im3(im3==i)=0; end end % figure(4); imshow(im3); title('denoised figure'); img_mask=imfill(im3, 'holes'); figure; imshow(img_mask); [imf NUM_SMA] = bwlabel(img_mask,4); %stats=regionprops(imf,'area','perimeter','centroid'); img_mask=label2rgb(imf,'cool','k','shuffle'); figure;imshow(img_mask); title(['number of cells =', num2str(NUM_SMA)]); output = NUM_SMA; function output = DAPI_count(image) img = imread(image); %figure(1); imshow(img); imgn = imresize(img(:,:,3),0.3); %figure(2);imshow(imgn); %level = graythresh(imgn); im = im2bw(imgn,.2); im2 = imclearborder(im); %clear the cells around the border %figure(3); imshow(im2); [im2 N]= bwlabel(im2,4); %Remove noise for i = 1:N Area = length(find(im2 == i)); if Area > 300 && Area < 50 im2(im2==i)=0; end end figure(4); imshow(im2); title('denoised figure'); [imf NUM_DAPI] = bwlabel(im2,4); img_mask=label2rgb(imf,'hsv','k','shuffle'); figure;imshow(img_mask); title(['number of cells =', num2str(NUM_DAPI)]); output = NUM_DAPI;

MATLAB code to calculate distances between patterns

%Code to automatically calculate the distances between patterned cells %Christine Ho, Sept 10, 2013

%Instructions %Add images into one folder named "Count", then run program function out = IntercellDistOneCh %Intercellular_dist should be file name clc, clear, close all sdir = 'Count1Ch'; %folder name

63 tif = dir([sdir '/*.tif']); %files names in the list Distances = []; %empty matrix of the output of the distances for i = 1:length(tif) %for as many images, run these commands image = [sdir '/' tif(i).name];% one image file at a time DIST_CELLS = Cell_dist(image); Distances = [Distances; DIST_CELLS]; end out = Distances function output = Cell_dist(image)

%image= imread('12 22 2012 1 to 10 6 hours_20x tile2.tif'); %split image into the two channels %imcancergreen=image(:,:,1); image= imread(image); figure; imshow(image);

%threshold levelg=.0569; %levelg=graythresh(image); new_green=im2bw(image,levelg); figure; imshow(new_green);

[L, N]=bwlabel(new_green); %regionprops gives the area & perimeter stats=regionprops(L,'area','perimeter', 'Centroid','MajorAxisLength','MinorAxisLength'); areas=[stats.Area]; perimeters=[stats.Perimeter]; too_small=find(areas<100); %find the areas that are too small to be cells img_mask2=L; for i=1:length(too_small) img_mask2(L==too_small(i))=0; end

%label the fibroblasts & cancer cells separately [Can Ngreen]=bwlabel(img_mask2); figure; imshow(Can); stat_Can = regionprops(Can, 'Centroid');

%D = zeros(Ngreen, numred); %row = dist btwn designated fibroblast and cancer cells; %column = different fibroblasts for n = 1:Ngreen pix = find(Ngreen==n); b2 = logical(0*Ngreen); %make entire pic zero b2(pix) = 1; %display only the pix in that one cell bp = bwperim(b2); %get the perimeter around cell

[r c]=find(bp); plot(c,r,'.r','markersize',3) %plot the dots around the perimeter

64

hold on centroid = cat(1,stat_Can.Centroid); text(centroid(n,1),centroid(n,2),{'\fontsize{15}\color{magenta}\bf',num2str(n)}); hold on end title(['number of cells =', num2str(Ngreen)]);

D=zeros(Ngreen, 1); for n2 = 1:Ngreen centroidONE = stat_Can(1).Centroid; centroidX = stat_Can(n2).Centroid; dx = centroidX(1)-centroidONE(1); dy = centroidX(2)-centroidONE(2); %dx2=POWER(dx,2); dist= sqrt(dx^2 + dy^2); D(n2,1) = dist; end