The Screening of Biomaterials to Support Long-Term Growth and Maintenance of Human Embryonic Stem Cells in Xeno- and Feeder-Free System

Transcription

1 The Screening of Biomaterials to Support Long-Term Growth and Maintenance of Human Embryonic Stem Cells in Xeno- and Feeder-Free System by Justin Tse Wei Pang A thesis submitted in conformity with the requirements for the degree of Master of Applied Science in Biomedical Engineering Institute of Biomaterials and Biomedical Engineering University of Toronto Copyright by Justin Tse Wei Pang 2013

2 The Screening of Biomaterials to Support Long-Term Growth and Maintenance of Human Embryonic Stem Cells in a Xeno- and Feeder-Free System Abstract Justin Tse Wei Pang Master of Applied Science in Biomedical Engineering Institute of Biomaterials and Biomedical Engineering University of Toronto 2013 Current feeder-free culture systems employing undefined Matrigel are still more effective in maintaining human embryonic stem (ES) cells than defined surfaces using extracellular matrix (ECM) proteins. While the role of substrate stiffness in stem cell fate is becoming increasingly evident, all previous culture systems use ECM proteins on rigid polystyrene surfaces. Here, we used factorial designs to screen and evaluate combinations ECM proteins and substrate stiffness for their effect on short-term pluripotency and self-renewal. Using optimal conditions determined from our screening experiments, defined and near xeno-free culture systems maintained CA1 human ES cells for over 10 passages in Essential 8 (E8) medium. Under these conditions, we found that human ES cell self-renewal was greater on soft polydimethylsiloxane (PDMS) substrates than on rigid polystyrene dishes. The culture systems and screening tools developed in this project will help develop robust and defined xeno-free culture systems that incorporate both biochemical and biomechanical factors. ii

3 Acknowledgments First and foremost, I would like to acknowledge and thank my supervisor Dr. Craig Simmons for all the support and advice I have received over the past 2 years. While my time here has been short, by now I already know I will probably never find another supervisor as cool as Craig. His down to earth demeanor, along with his passion and enthusiasm towards research sets a positive atmosphere in the lab that I will miss. None of the work that I have accomplished here would have been possible without Craig s ability to never be fazed whenever I brought forth seemingly impossible obstacles and roadblocks. He would offer smart ideas here and there, a little pinch of encouragement, and I was on my way with newfound inspiration. In the end, I am truly grateful for the opportunity to work in your lab and on a project that I thoroughly enjoyed. I wish you and your family all the best. I would also like to thank in particular, Mark and the Dr. Lye Lab for teaching me all that I know about stem cells; Kelly for showing me the ropes to the Simmons lab, statistical design, and fixing the protein printer; Haijaio for helping me with your favourite toy, the AFM; and lastly Zahra for general technical support and scientific shopping. Thank you all so much for enduring and answering the same questions multiple times until I finally got the sense to write it down. To everyone else in the lab and former members, far too many amazing people to list, thanks for reciprocating to all nonsense I loved to talk about. I will definitely miss all the lunches, dinners, and night outs we had as a lab, as well as all the in-lab discussions, it has been hilarious. Thanks to you all, I have nothing but positive memories and thoughts from my time here. Lastly, I would like to thank my friends and family for reminding me there happens to be a life outside my studies and the internet. I would especially like to thank my mother and grandmother for everything they have done for me, as their continued support has allowed me achieve my goals and dreams. iii

4 Table of Contents Acknowledgments... iii Table of Contents... iv List of Tables... vii List of Figures... viii List of Appendices... xi Chapter 1 Introduction Motivation...1 Chapter 2 Background Introduction Core Transcriptional Factors Culturing Human ES Cells The Standard Culture System Human Feeders Layers Matrigel/Geltrex Beginning to Define the Culture System Supplementing Growth Factors into ES media Defined Culture Substrates Matrix Stiffness Statistical Design and Analysis of Experiment Introduction k Factorial Designs Calculating Main Effects and Interactions Center Points and Fractional Factorial Designs Blocking and Fractional 2-Level Factorial Designs...40 iv

5 Chapter 3 Chapter 4 Thesis Hypothesis and Objectives...44 Overview of Experimental Designs and Methods Measurement of the feeder layer stiffness by AFM Creating physiologically stiffness relevant polydimethylsiloxane (PDMS) substrates Feeder and feeder-free human ES cell cultures Experimental design to screen for pluripotency and self-renewal of human ES cells Validating optimal and suboptimal substrate conditions determined from the screening experiments ALP staining and immunostaining...51 Chapter 5 Results Feeder layer stiffness measured by AFM The significant factors in pluripotency and self-renewal of human ES cell cultures determined from the screening experiments Significant factors of human ES cells grown mtesr Optimal substrate conditions for human ES cells in mtesr Significant factors of human ES cells grown E Optimal substrate conditions for human ES cells in E Suboptimal conditions defined by screens Validating optimal and suboptimal substrate conditions determined from the screening experiments CA1 cultures on optimal substrate conditions grown in mtesr CA1 cultures on optimal substrate conditions grown in E CA1 cultures on the suboptimal substrate condition...73 Chapter 6 Chapter 7 Discussion...75 Conclusions and Future Work Conclusions Future Work...83 v

6 7.2.1 Characterization of PDMS substrates Further characterization of new and current hit conditions...84 References...85 Appendix...97 vi

7 List of Tables Table 2-1: Summary of successful human ES culture systems on defined ECM proteins in nonconditioned medium Table 2-2: An example ANOVA of a statistically significant regression model Table 2-3: Table of contrasts for a 2 3 factorial design Table 2-4: 2 3 factorial design augmented with 3 center points Table 2-5: ABCDE+ block of a 2 5 factorial design Table 2-6: 2 5 factorial design partitioned into the signs of ABCD and BCE Table 5-1: Estimated regression coefficients of simplified fitted model for ALP staining intensity and Oct4+ cell count in mtesr1 medium Table 5-2: Estimated regression coefficients of simplified fitted model for ALP staining intensity and Oct4+ cell count in E8 medium Table 5-3: Summary of substrate conditions vii

8 List of Figures Figure 2-1: The core transcriptional regulatory circuitry of ES cells. All three core transcription factors interact with each other in the form of positive feedback loops. They co-occupy many additional genes necessary for self-renewal and pluripotency. At the same time, they repress the expression of many important developmental genes required for lineage commitment. Reprinted by permission from Macmillan Publishers Ltd: Nature Reviews Molecular Cell Biology (67), copyright Figure 2-2: Model showing the role of Nanog in maintaining human ES cells. Nanog is upregulated by activin/nodal, which blocks neuroectoderm differentiation by FGF signaling. Nanog also inhibits Smad2/3 in order to reduce endoderm specification by activin/nodal signaling. (Source: Vallier et al. (102) ) Figure 2-3: Oct4 expression and ALP staining of mouse ES cells with or without LIF on PA gel versus rigid dish. A) High expression of Oct4 and ALP in mouse ES cells on 0.6 kpa gel in LIF+ medium. B) High expression of Oct4 and ALP in mouse ES cells on rigid dish in LIF+ medium, but arrows point to the presence of differentiated colony. C) High expression of Oct4 and ALP in most mouse ES cells on 0.6 kpa gel in LIF- medium. D) Differentiated mouse ES cells on rigid dish in LIF- medium. (Source: Adapted from Chowdhury et al. (132) ) Figure 2-4: Comparison of experimental design between OFAT experiment and factorial design of an example 3 factor experiment Figure 2-5: Geometric view of the 2 3 factorial design with hypothetical observations Figure 2-6: Significant lack of fit using a first order linear model, but not with a second-order quadratic model Figure 5-1: Stiffness of various XHEF, XHFF, and MEF feeder layers. The error bars indicate standard error. Statistical significance of P < Figure 5-2: Response surface maps of ALP staining intensity in mtesr Figure 5-3: Response surface maps of Oct4+ cell count in mtesr viii

9 Figure 5-4: Response surface maps of ALP staining intensity model in E Figure 5-5: Response surface maps of Oct4+ cell count model in E Figure 5-6: Response surface maps of demonstrating less optimal substrate conditions determined from different models Figure 5-7: Validation results of different substrate conditions either in mtesr1 or E8. The cultures that were terminated because of poor attachment and survival are marked with a X. Some substrates were tested multiple times, and they are denoted by a second or third darker shaded bar. The arrows indicate the cultures are currently still being maintained Figure 5-8: Growth curves of CA1 grown on various substrate conditions in E8. Cells were grown for 5-6 days before being counted and passaged onto new substrates. Each culture and passage was started with cells Figure 5-9: Growth curves of CA1 grown on various substrate conditions in mtesr1. Cells were grown for 8 days before being counted and passaged onto new substrates. Each culture and passage was started with cells Figure 5-10: Percentage of cells expressing Oct4 or Nanog pluripotency marker in CA1 cultures on different substrate conditions in mtesr1. Most substrate conditions in mtesr1 did not survive to passage 5 to access their pluripotency. Oct4 and Nanog percent expression was determined through immunostaining and image analysis Figure 5-11: Percentage of cells expressing of Oct4 or Nanog pluripotency marker in CA1 cultures on different substrate conditions in E8. Substrate condition 2 on plastic did not survive up to passage 5 to have its pluripotency markers reassessed. Oct4 and Nanog percent expression was determined through immunostaining and image analysis Figure 5-12: Representative phase contrast images of CA1 cells on substrate condition 1 in mtesr1 after five days for passage 1 and 2. On plastic, cell numbers were comparable between passage 1 and 2, but on 3 kpa substrates, cell numbers were diminished in passage 2. Scale bars, 100 µm ix

10 Figure 5-13: Representative phase contrast images of CA1 cells on different substrates in mtesr1 stained with ALP. Scale bar, 100 µm Figure 5-14: Representative phase contrast images of CA1 cells on various substrate conditions in E8 after five days. While CA1 cells overall grew faster than cultures in mtesr1, substrate conditions on soft substrates had even more robust growth. Scale bars, 100 µm Figure 5-15: Representative phase contrast images of CA1 cells on different substrates in E8 stained with ALP. Scale bar, 100 µm Figure 5-16: Phase contrast images of CA1 cells on the suboptimal substrate condition in mtesr1 on plastic. While CA1 cells were able to be maintained to the third passage, they began to detach after day 4 and the culture was lost. Scale bars, 100 µm x

11 List of Appendices Table A1: The engineering and coded values of each factor used in the factorial design Table A2: The 19 experimental runs of a fractional factorial design with three center points. The runs were ordered by stiffness for clarity, but the runs were performed randomly Figure A1: JMP report of the screening experiment performed in mtesr1 and measured by ALP staining intensity Figure A2: JMP report of the screening experiment performed in mtesr1 and measured by Oct4+ cell count Figure A3: JMP report of the screening experiment performed in E8 and measured by ALP staining intensity Figure A4: JMP report of the screening experiment performed in E8 and measured by Oct4+ cell count Figure A5: Varying amounts of fluorescent protein were incubated overnight on plasma treated PDMS substrates and washed twice with PBS. More adsorbed fibronectin was observed through increased fluorescence as the incubated protein density increased and PDMS stiffness was lowered. In contrast, the amount of adsorbed laminin generally increased with the concentration applied (with weak sensitivity), but appeared to be independent of stiffness xi

12 1 Chapter 1 Introduction 1.1 Motivation In 1998, human embryonic stem (ES) cell lines were first established from embryos produced by in vitro fertilization (1). The embryos were grown in culture to the blastocyst stage and the inner cell mass was isolated to generate human ES cells. These cells are characterized by their ability to self-renew indefinitely and their capacity to differentiate into derivatives of all three germ layers (2). Their potential for self-renewal and differentiation into any cell type, if cultured appropriately, has great potential for cell and tissue transplantation in treating many diseases (3). Currently, donated organs and tissues are used to replace damaged ones, but the demand for organs far exceeds the available supply, leaving a void that can be filled with stem cell therapy. Recent advances in human ES cell research has shown potential in treating diseases including amyotrophic lateral sclerosis (ALS), Alzheimer s disease, Parkinson s disease, epilepsy, stroke, spinal cord injuries, blindness, diabetes, heart disease, tissue/skin damage, cancer, muscular dystrophy, hearing loss, and others (4-8). Human ES cells have previously been directed into several functional cell types including neurons secreting dopamine, beta-cells secreting insulin, and cardiomyocytes responding to electrical signals (9-13). There are active clinical trials using human ES-derived retinal cells to treat patients with Stargardt s Macular Dystrophy and agerelated macular degeneration that show preliminary improvement in vision without negative side effects (14). In addition to their uses in regenerative medicine, human ES cells will also be important in the development of new drug discovery platforms and as an in vitro research tool to study human development (15). It is expected that understanding how human ES cells behave and discovering ways to manipulate them will lay the groundwork for the future of personalized medicine using patient specific induced pluripotent stem cells (IPSCs) (16). However, safety is still a major issue that is preventing ES cells and IPSCs from becoming a viable therapy in regenerative medicine. A significant amount of the concern lies with today s human ES cell culture methodology because of the use of animal products and undefined materials. For researchers developing clinical therapies, it is important to use defined materials and to avoid animal products to comply with strict regulatory rules. It is also of interest to those studying stem cell biology as it is difficult to discover influences on signaling pathways if

13 2 undefined proteins or small molecules are present in the culture (17). Current methods used to establish and maintain self-renewing and pluripotent human ES cells involve the use of mouse embryonic fibroblasts (MEFs) feeder layers. Originally, they were used to support the selfrenewal and pluripotency of mouse embryonic stem cell (mescs), but have since been extended to human ES cells and to other stem cells (1, 18, 19). To be specific, when co-cultured with MEFs, human ES cells maintain long-term pluripotency, but subsequent removal of the feeder layer leads to spontaneous differentiation into many cell types (1, 15). Using xenogenic feeder layers such as MEFs adds additional risks such as infection and transplantation rejection, on top of tumorigenicity for patients (20). Mouse cells may carry pathogenic bacteria, viruses, and parasites that can contaminate human ES cells being cocultured with MEFs (15). Moreover, animal products such as MEFs are sources of nonhuman sialic acid Neu5Gc, which human cells are able to incorporate into their membranes (21). This can lead to an immune response and rejection of human ES cell derived tissues as most humans have antibodies specific for Neu5Gc. The possibility of xenogenic contamination would restrict all future clinical applications, and is one of many major hurdles in regenerative medicine and stem cell based therapies. Newer xeno-free methods, including the use of human foreskin fibroblast or embryonic fibroblasts as feeders to derive and support long term cultures of human ES cells have been recently demonstrated (22-24). However, feeder layers including MEFs have been shown to exhibit significant batch to batch variability that affects human ES cell self-renewal and differentiation (25). The use of feeder layers also suffers from being labour intensive and costly, restricting the large scale production of hescs and its derivatives. One potential solution would be to develop a chemically defined, xeno- and feeder-free culture system. This would avoid the risk of xenogenic contamination, and create more reproducible stem cell cultures while lowering the cost of production. These advances would help produce a near limitless source of human ES cells for deriving specific cell types required for regenerative medicine (25). Recent work has demonstrated that defined xeno- and feeder-free culture systems using extracellular matrix (ECM) proteins on polystyrene dishes can maintain pluripotency and self-renewal of human ES cells (26-28). However, such systems are not as effective in supporting and propagating long-term cultures of human ES cells compared to traditional feeder-dependent systems. Here, we propose to design a culture system that mimics the microenvironment of a feeder layer using combinations of ECM proteins and substrate stiffness. Factorial designs will

14 3 be used to determine optimal substrate conditions that will robustly promote pluripotency and self-renewal. It is expected that these approaches will bring many clinical applications of differentiated human ES cells and IPSCs closer to reality.

15 4 Chapter 2 Background 2.1 Introduction Human embryonic stem cells were first derived in 1998 from pre-implantation blastocyst (1). These cells have the ability to sustain prolonged undifferentiated proliferation as well as the developmental potential to form derivatives from all three germ layers even after long-term culture. The derivation of human ES cell lines enables the opportunity to study early development as well as holding great potential for regenerative medicine (1, 29). Human ES cells characteristically have a high nucleus to cytoplasm ratio, prominent nucleoli, and form distinct cell colonies. They are also known to express high levels of telomerase, which helps maintain telomere length and is correlated with cell line immortality (30). Much of our understanding of the transcriptional regulatory circuitry behind pluripotency and self-renewal of human ES cells originated from studying mouse ES cells (31). Mouse ES cells have been established in culture many years prior to human ES cells and mouse ES cells do not raise ethical issues of having to destroy human embryos (32). Moreover, mouse ES cells can be cultured in their undifferentiated state without feeder layers by supplementing leukemia inhibitory factor (LIF) into the culture medium (33, 34). LIF is a cytokine that signals through a heterodimeric receptor containing LIF receptor (LIFR) and gp130 that enables the maintenance of pluripotent mouse ES cells (35). The activation of Stat3, the key transcription factor downstream of the LIF/gp130 receptor complex is responsible for the underlying pluripotency. Induced expression of a dominant interfering (negative) Stat3 mutant in mouse ES cells resulted in differentiation even in the presence of LIF (36). LIF dependent suppression of differentiation is also eliminated with a point mutation of the tyrosine residue of gp130 implicated for Stat3 binding (37). However, human ES cells differ significantly from mouse ES cells in morphology and cell surface marker expression, along with their non-responsiveness to LIF (1, 38). Human ES cells instead are instead supported by members of the fibroblast growth factor (FGF) and transforming growth factor beta (TGF-β) family. As a result, one cannot assume that the human equivalent regulatory transcription factor will have the same function as it would in mouse ES cells.

16 5 2.2 Core Transcriptional Factors Human ES cells are regulated by three core transcription factors: Oct4, Sox2, and Nanog (39). These transcription factors play important roles in early development, but are also necessary for in vitro propagation of undifferentiated human ES cells. Both Oct4 and Nanog are homeodomain transcription factors that are evolutionarily conserved to play a role in cell-fate specification. Homeodomain proteins regulate anatomical development in animals, fungi, and plants by activating networks of transcription factors and signaling molecules (40). While they can be involved in maintaining pluripotency (Oct4 and Nanog are examples), homeodomain transcription factors can also specify cell identity through differentiation. For example, Hox genes encode for a homeodomain containing proteins that are able to direct the embryonic identity of body segments along the antero-posterior axis in both vertebrates and invertebrates. Conversely, Oct4 promotes pluripotency and is encoded by Pou5f1 that contains a POU domain, which binds with an octamer sequence found in promoters and enhancers of many genes (41). Oct4 is known to be an essential regulator of early development and ES cell identity that is expressed early in embryonic development during the four cell stage. Once the blastocyst is formed, its expression is restricted to the inner cell mass (ICM), which is a mass of pluripotent stem cells that will give rise to the embryo (42). The outer layer of cells, known as the trophectoderm will give rise to the placenta, and is devoid of Oct4 (43). Mutant mouse embryos that are Oct4-deficient die shortly after implantation (42). These embryos are able to proliferate and survive up to mid-blastocyst stage, but are unable to form well defined ICMs and do not survive. Oct4-deficient blastocysts lack a discernable embryonic compartment after implantation because the cells of the ICM are not pluripotent. In vitro, cultures of Oct4-deficient blastocysts survive, but only generate trophoblast cells (42). As a result, ES cells cannot be derived from embryos lacking Oct4, suggesting that Oct4 is essential for pluripotency. One study using a tetracycline-regulated Oct4 transgene to control levels of Oct4 expression in mouse ES cells established a role for Oct4 as a master regulator of pluripotency (44). To maintain a pluripotent state, ES cells must keep Oct4 tightly regulated at a certain concentration. If the level of Oct4 is increased by twofold, ES cells are directed towards differentiation into primitive endoderm and mesoderm. However, if Oct4 is diminished, ES cells are triggered towards differentiation into trophectoderm. In the case of increased Oct4, upregulation of Gata4 and

17 6 brachyury mrna, primitive endoderm and mesoderm markers respectively, was evident. Conversely, a reduction in Oct4 led to an upregulation of Cdx2 and Hand1 mrnas, transcription factors involved in trophoblast differentiation. No primitive endoderm or mesoderm markers were detected in this instance. As a result, Oct4 may also have an additional role in restricting ES cells from dedifferentiating into the trophectoderm lineage. Further evidence on Oct4 s importance in pluripotency is observed using small interfering RNA (sirna) to knockdown Oct4 in both mouse and human ES cells (45, 46). Disruption of Oct4 in mouse ES cells leads to flatten differentiated cells within 48 hours. The cells continue to proliferate and look morphologically similar to mouse ES cells culture after the withdrawal of LIF. Without Oct4, the expression of Cdx2, a trophectoderm marker, along with some endoderm markers, Gata6 and α-fetoprotein (AFP) was observed. Neither brachyury (mesoderm marker), Pax6 (ectoderm marker), nor FGF5 (primitive ectoderm marker) was detected with Oct4 knockdown. The upregulation of Cdx2 is in agreement with the results from the tetracyclineregulated Oct4 transgene, but increased expression of endoderm genes was also observed (44, 45). Similarly with human ES cells, knockdown of Oct4 led to flat and slower growing colonies that were indicative of differentiation (46). Primitive endoderm markers such as Gata6 and Gata4, as well as trophoblastic markers human chorionic gonadotrophin alpha (hcgα) and beta were upregulated following the knockdown of Oct4 (45, 47). Unlike mouse ES cells, induction of other endoderm markers such as AFP or trophoblast differentiation markers, Gata2 or Cdx2 was not immediately detected under culture conditions that support long-term self-renewal of human ES cells. Only when suboptimal conditions were used were the additional endoderm and trophoblast differentiation markers seen. No significant changes to several mesodermal or ectodermal markers were observed with Oct4 reduction (45, 47). In addition to monitoring gene expression, the expression of cell surface antigens can also be telling of differentiation. Human ES cells express several cell surface markers including Stage Specific Embryonic Antigen-3 (SSEA3), SSEA4, TRA-1-60, and TRA-1-81, but not SSEA1 (38). It has been shown that human ES cells in culture express the same cell surface markers listed previously as the ICM cells from human blastocysts (48). Moreover, human ES cells express the liver/bone/kidney isozyme of alkaline phosphatase (AP), which can be detected by antibody TRA In this regard, human ES cells are similar to human embryonal carcinoma (EC) cells, but different compared to mouse ES and EC cells, which express SSEA1 and AP, but not

18 7 SSEA3, SSEA4, TRA-1-60-, or TRA-1-81 (48). When human ES cells are induced towards differentiation, whether by removing MEF feeder layer or with retinoic acid, the expression of SSEA3, SSEA4, TRA-1-60, TRA-1-81, and AP (TRA-2-54) was down regulated (38). The speed of disappearance during differentiation varies among the antigens. SSEA3, TRA-1-60, TRA-1-81 was lost quickly, while AP disappeared relatively slowly, and SSEA4 was lost at an intermediate speed. The loss of these pluripotent markers is accompanied by changes in cell morphology indicative of differentiation. Likewise, when Oct4 expression is reduced by sirna, the expression of SSEA3, SSEA4, TRA-1-60, and TRA-1-81 are all significantly reduced, alongside changes in culture morphology and growth rate typical of cells undergoing differentiation (46, 47, 49). Some cells positive for SSEA1, which is not normally expressed by human ES cells, was also found in culture following Oct4 knockdown. This is suggestive of differentiation into trophectoderm as SSEA1 is strongly expressed in human trophectoderm, but not in human ICM (48, 49). Overall, Oct4 is required to maintain the undifferentiated stem state and the loss of Oct4 results in differentiation. Another homeodomain transcription factor that is important for the maintenance of embryonic stem cells is Nanog. First discovered in mouse, its expression is restricted to pluripotent cells (50). Its mrna is highly expressed in ES cells, embryonic germ (EG) cells, and EC cells, but not in more differentiated cells in the parietal endoderm, hematopoietic stem cell lines, or any adult tissues. Its expression is limited to the ICM of mouse preimplantation blastocyst, and the loss of Nanog prevents the formation of the epiblast, resulting in early embryonic lethality (51). Constitutive expression of Nanog in mouse ES cells maintains the undifferentiated state in prolonged cultures without LIF/Stat3 activation. Elevated Nanog works independently of LIF/Stat3 to promote ES cell self-renewal, cell growth, and to maintain Oct4 levels (50, 51). However, Oct4 is still essential for pluripotency as the loss of Oct4 results in differentiation even with continued expression of Nanog. In human preimplantation embryos, Nanog is not detected in early cleavage stages, nor in the early morulae, but restricted to the ICM of the blastocyst (52). Its restricted expression is very similar to that of Nanog in mouse embryos (50). In culture, when Nanog is downregulated with sirna, human ES cells show morphology indicative of differentiation as they flatten and enlarge within 2 days. The knockdown of Nanog also reduces the expression of Oct4, and a reciprocal effect is true where RNA interference (RNAi) of Oct4 also leads to a reduction of Nanog

19 8 expression (47). This effect is observed early on as differentiation is just starting, suggesting an interplay of gene regulation between the two core transcription factors. Human ES cells also show greatly reduced cell surface markers including SSEA3, SSEA4, TRA-1-60 and TRA-1-81 following Nanog knockdown with sirna, indicating the loss of pluripotency (47, 49, 52). They appear to differentiate down the primitive endoderm and trophectoderm lineages, but not ectoderm or mesoderm lineages as levels of FGF5, PAX6, and brachyury show no difference between cells transfected with Nanog or control sirna. However, trophoblast markers, hcgα, hcgβ, CDX2, GATA2 SSEA1, and primitive endoderm markers, GATA4, GATA6, laminin B1, AFP are all upregulated compared to control cells. This is consistent with Nanog s reported role in mouse embryos and ES cells as a suppressor of primitive endoderm differentiation (50, 51). However unlike in mouse, differentiation towards trophectoderm also occurs and thus in humans, Nanog may also serve to repress trophectoderm differentiation. The expression of Nanog in ES cells has been shown to be controlled by the two other core transcription factors essential to maintaining pluripotency, Oct4 and Sox2 (53). Within the Nanog proximal promoter, the highly conserved 15 base pair sequence identifying the composite soxoct cis-regulatory element was found in both humans and mouse ES cells. The sox-oct enhancer element contains the Sox2 and Oct4 binding site, important for the cooperative interaction of Sox2 and Oct4 that have been previously shown to upregulate other pluripotent associated genes. More importantly, Oct4 and Sox2 have been shown by chromatin immunoprecipitation (ChIP) to bind to the Nanog promoter in both mouse and human ES cells. Yet, a recent study showed that at least in mouse ES cells, Nanog expression seems to fluctuate in ES cells, where a fraction of Oct4 + cells do not express Nanog (54). Nanog-GFP + and Nanog-GFP ES cells sorted by fluorescence-activated cell sorting (FACS) both redistribute into mixed populations upon reculturing. Although cells not expressing Nanog are able to remain undifferentiated and can reexpress Nanog, they grow slowly and give rise to fewer undifferentiated ES colonies than Nanog expressing cells. It turns out that while Nanog may not be strictly required to maintain pluripotency as Nanog null cells maintained the expression of all embryonic stem-cell associated transcripts (ECATs) except Nanog itself, they show lower self-renewal capacity and lower resistance towards differentiation (51, 54). This is consistent with the increased resistance towards differentiation with Nanog overexpression (50, 51). Therefore the loss of Nanog makes the cells

20 9 more prone to differentiate, but it appears to not commit the cell s fate towards differentiation (54). The third core transcriptional factor that is central to the transcriptional regulatory hierarchy of ES cells is Sox2. It belongs to the Sox gene family that encodes for transcription factors containing a single DNA binding HMG-box protein domain (high mobility group box) (55). Like Oct4, Sox2 is present in the ICM, epiblast and germ cells. However, unlike Oct4, Sox2 is also detected in some multipotent embryonic and extraembryonic lineages including neuroectoderm and in the ectoplacental cone respectively. Mouse embryos deficient in Sox2 are lethal shortly after implantation. In fact, null mutant blastocysts grown in culture do not develop a normal ICM or epiblast, and Oct4 expression is quickly lost in these cultures. Similarly to Oct4 null mutants, when the ICM of Sox2 null mutants is excised from the blastocyst and grown in culture they form trophoblast cells instead of epiblast (42, 55). Furthermore, in culture, mouse ES cells that are deficient of Sox2 through a tetracycline regulated Sox2 transgene, differentiated into mostly trophectoderm cells similar to Oct4-null ES cells (44, 56). These cells stained strongly for the trophectoderm markers cadherin 3 and cytokeratin 7, while stained very weakly for mesoderm and extraembryonic endoderm cells (57). Likewise, when sirnas are used to knockdown Sox2 mrna in human ES cells, stem cell antigens SSEA3, SSEA4, TRA1-60, and TRA1-81 is reduced in culture and differentiation into trophectoderm occurs (49). Human trophectoderm markers including SSEA1, human placental lactogen, and cytokeratin 7 are observed. In addition, the knockdown of Sox2 leads to significant reduction of Nanog and Oct4 expression levels in 4-6 days. Similar to the Oct4 and Nanog reciprocity effect mentioned earlier, a reciprocal effect also occurs here, where the knockdown of either Nanog or Oct4 results in a drastic reduction of Sox2 levels. Sox2 and Oct4 have been previously implicated in regulating expression of several pluripotency associated genes through enhancer elements containing Sox2 and Oct4 binding motifs, including Fgf4, Lefty1, Utf1, Fbx15, and as mentioned, Nanog (53, 58-61). Oct4 and Sox2 are able to independently bind to their respective DNA binding site and their interaction activates downstream transcription of pluripotent genes. This is supported by crystal structures of a POU/HMG ternary complex bound to composite sox-oct element (62). This is further supported by the observation that transient transfection of the FGF4 enhancer on chloramphenicol acetyltransferase (CAT) into differentiated cells, only shows expression of the reporter when

21 10 both Sox2 and Oct4 expression plasmids are also transfected (58). The transfection of just one of two expression plasmid or with one plasmid in conjunction with Oct1 failed to promote CAT gene expression. Mutations or deletion of either binding motif in the enhancer also dramatically reduces the level of expression of specific gene such as Lefty1 or Fgf4 in ES cells (59). Moreover, this Sox2-Oct4 binding element has also been found in enhancers regulating both Sox2 and Oct4 themselves. Using ChIP of mouse and human ES cells, both Oct4 and Sox2 were shown to bind directly to the enhancers of genes encoding Oct4 and Sox2, suggesting a positive feedback loop (63). The sequence of the sox-oct composite element has been found to be highly conserved between species, including mouse and humans (59). Of the pluripotency associated genes mentioned, Fgf4, Lefty1, Utf1, Fbx15, Nanog, Oct4, and Sox2 all share near identical sox-oct sequence between mouse and human genes. The evidence suggests that Oct4 and Sox2 are able to cooperatively regulate the transcription of target genes including themselves and Nanog, as well as co-occupy many downstream genes. Using ChIP followed by DNA microarrays, it was shown that Oct4, Sox2, and Nanog are associated with the promoters of hundreds of protein-coding genes and several small noncoding RNA called microrna (mirna) genes in human ES cells (39). Not surprisingly, about half the genes bound by Oct4 are also bound by Sox2. Interestingly, greater than 90% of genes cooccupied by both Oct4 and Sox2 are bound by Nanog. At least 353 of annotated human genes are occupied by all three transcription factors, and approximately half have their transcript consistently detected in human ES cells. Notable active genes include the transcription factors Oct4, Sox2, Nanog, Stat3, and components of the TGF-β and Wnt signaling pathways that have been implicated in pluripotency and self-renewal in human and mouse ES cells (64, 65). Similarly, many of the inactive genes co-occupied by Oct4, Sox2, and Nanog where no transcripts are detected are transcriptional factors necessary for differentiation into extra-embryonic, ectodermal, endodermal, and mesodermal lineages. Many of these inactive genes code for the developmentally important homeodomain proteins that are important in specifying cell identity (39). Given that Oct4, Sox2, and Nanog are all downregulated upon differentiation, it appears that they interact with these genes in a regulatory fashion. To support this, when the expression levels of genes co-occupied by Oct4, Sox2, and Nanog in human ES cells was compared with 79 human differentiated cell types, a significant difference in expression was detected (39, 66). Thus, it suggests that Oct4, Sox2, and Nanog are activating co-occupied genes found transcriptionally

22 11 active in ES cells while repressing the co-occupied inactive genes in order to maintain pluripotency and self-renewal. The loss of these regulators during differentiation allows the upregulation of genes important for development and for the reduction of gene expression to maintain the stem cell state (Figure 2-1). Figure 2-1: The core transcriptional regulatory circuitry of ES cells. All three core transcription factors interact with each other in the form of positive feedback loops. They co-occupy many additional genes necessary for self-renewal and pluripotency. At the same time, they repress the expression of many important developmental genes required for lineage commitment. Reprinted by permission from Macmillan Publishers Ltd: Nature Reviews Molecular Cell Biology (67), copyright Surprisingly, despite Sox2 s occupation with many genes necessary for maintenance of pluripotency and self-renewal, at least in mouse ES cells, Sox2 has been shown to be dispensable for the activation of Oct-Sox enhancers (56). While repression of Sox2 leads to differentiation to trophectoderm, the expression of previously mentioned pluripotent stem cell-specific genes Fgf4, Lefty1, Utf1, Nanog, and Oct4 are all downregulated relatively slowly compared to the repression of Oct4. To study this further, luciferase reporter constructs containing the wild type

23 12 oct-sox enhancers and mutant sox site of Fgf4, Lefty1, Utf1, Nanog, Oct4, and Sox2 were transfected into ES cells with the inducible Sox2 repression system. After inducing repression of Sox2, the expression of these genes containing wild type oct-sox enhancers is unchanged from uninduced cells for 24 to 48 hours before expression rapidly falling. For the sox mutant binding site, there was reduced expression in both induced and uninduced cells. The results suggest that Sox2 may not be required for activation of these sox-oct constructs and that other Sox factors may redundantly contribute to the activation of these enhancers. Indeed other Sox transcription factors such as Sox15, Sox11, and Sox4 are expressed in ES cells, and have been reported to be capable of interacting with the oct-sox enhancer element (68-70). ChIP was then used to show that these factors were bound to the oct-sox enhancers of several pluripotent stem cell specific genes, indicating that several Sox factors can activate oct-sox enhancers in ES cells, making Sox2 dispensable in this function (56). More importantly, it is thought that Sox2 may have a role in regulating Oct4 expression. Using DNA microarrays to screen for immediate changes in gene expression following Sox2 reduction revealed several genes known for regulating Oct4 expression to be differentially expressed (56). One of those genes, Nr5a2, which encodes for the nuclear receptor LRH-1that helps maintain Oct4 expression was downregulated to 15% after 24 hours (71). Conversely at the same time point, Nr2f2, which encodes for a negative regulator of Oct4, along with trophectoderm differentiation inducers such as Eomes and Esx1 are up-regulated to 170%, 260%, and 190% respectively (57). As a result, the changes in expression levels immediately following Sox2 repression should reduce Oct4 levels and promote trophectoderm differentiation. This was further validated by using an Oct4 transgene to maintain pluripotency and self-renewal of Sox2- null ES cells (56). Floxed Sox2 alleles were deleted by dexamethasone (Dex) inducible Cre recombinase and either a Nanog or Oct4 transgene was added. Only the Oct4, not Nanog transgene could rescue Sox2 knockout cells. Quantitative RT-PCR revealed that following Sox2 deletion, Nr5a2 is downregulated, along with the up-regulation of Nr2f2, Eomes, and Esx1. Consequently, endogenous Oct4 mrna levels accounted for less than half the Oct4, but exogenous Oct4 was able to rescue the Sox2 knockout ES cells. This suggests that at least in mouse ES cells, the main function of Sox2 is to mediate pluripotency by regulating Oct4 expression as opposed to activating oct-sox enhancers.

24 13 Since the discovery of human ipscs and mouse ipscs, a few more key regulators have been implicated in regulating the three core transcription factors. One of those regulators, Lin28, is a repressor of mirna processing and a post-transcriptional mrna regulatory factor (72). Lin28 was one of the four factors alongside Oct4, Sox2, and Nanog to be sufficient in reprograming human somatic cells to pluripotent stem cells (73). Qiu et al. revealed that Lin28 together with an interacting partner, RNA helicase A (RHA) regulates Oct4 expression post-transcriptionally (72). Lin28 has previously been shown to stimulate translation of specific target mrna in other cell types. In human ES cells, Lin28 appears to bind to specifically to Oct4 mrna and not to Sox2 or Nanog mrna as shown in immunoprecipitation (IP) experiments. Knockdown of Lin28 with sirna reduces Oct4 protein levels. In addition, knockdown of RHA reduces Lin28-stimulated translation of a luciferase reporter system containing the Lin28 binding region of Oct4 mrna. The evidence suggests that Lin28 regulates expression of Oct4 post-transcriptionally in human ES cells. Lastly, KLF4, a Kruppel-like zinc finger transcription factor and PBX1 (pre-b cell leukemia), a homeodomain transcription factor has been shown to regulate Nanog expression (74). KLF4 was one of the four factors alongside Oct4, Sox2, and c-myc to be sufficient in reprogramming mouse adult fibroblast into pluripotent stem cells (75). In human ES cells, high levels of KLF4 and PBX1 were detected, while upon differentiation, KLF4 and PBX1 were downregulated. Using a Nanog promoter luciferase reporter, Chan et al., showed that overexpression of KLF4 and PBX1 increased the expression of the Nanog promoter activity, as well as endogenous Nanog protein expression (74). Using ChIP, KLF4 was bound directly on the Nanog promoter, while PBX1 was bound to the promoter and an upstream enhancer. Along with Lin28, KLF4 and PBX1 expand the core transcriptional regulatory circuitry in human ES cell beyond the three core transcription factors. 2.3 Culturing Human ES Cells The Standard Culture System Human ES cells were first derived and cultured on inactivated MEF feeder layers with medium containing 10-20% fetal bovine serum (FBS) (1, 29). The feeder layers laid down extracellular matrix proteins (ECM) and secreted many growth factors suitable for pluripotency and proliferation. Grown on plastic in the absence of MEF feeder layers, human ES cells differentiate

25 14 regardless if LIF is supplemented in the culture (1). FBS has typically been used to support mammalian cultures and contains many proteins, but its complete composition is unknown and significant variability exists among batches (15). Moreover, the inclusion of FBS leads to extremely poor single cell cloning efficiency, suggesting the presence of chemicals that negatively affect stem cell survival. More importantly, as a consequence, the original derivation methods propagated cells as small clumps instead of single cells (1, 29). Thus, a small possibility existed that no single cell in the ES culture was able to create derivatives from all three germ layers, but there were multiple precursor stem cells committed to different lineages in the population. In order to improve the cloning efficiency of human ES cells for single cell derivation, a serum-free medium (KnockOut Serum Replacement, SR, Invitrogen), supplemented with 4-10 ng/ml basic fibroblast growth factor (bfgf) in KnockOut Dulbecco s modified Eagle s medium (DMEM) is used instead of FBS (76). SR contains insulin, transferrin, and bovine serum albumin (BSA), but supplementing bfgf in SR not only improves cloning efficiency, but it is necessary for continued pluripotency and self-renewal of human ES cells. Several human ES cell lines were clonally derived in these conditions, and were shown to self-renew, express high levels of telomerase and maintained normal karyotypes after 8 months in culture (76). As a result, for the standard culture system FBS was first replaced with 20% SR and 4-10 ng/ml bfgf on MEF feeder layers to support long-term growth and maintenance of human ES cells. (76, 77). Removing FBS and using SR in human ES cultures helps to standardize culture conditions as it is more defined and varies less between lot to lot compared to FBS. However, despite the improvement of cloning efficiency with SR, culture conditions need to continue to improve as the cloning efficiency of human ES cells is still less than 1%, which is less than that of mouse ES cells. Moreover, SR still contains undefined animal products, namely BSA. Regardless, SR supplemented ES medium cannot maintain undifferentiated human ES cells without MEF feeder layers or in MEF conditioned medium (MEF-CM), which are highly undefined and not suitable for clinical applications Human Feeders Layers Alternatively, human feeder layers instead of MEFs can be used to establish xeno-free cultures of human ES cells, including the derivation of clinical grade lines (22, 23, 78). Human ES cells derived from MEFs were first shown to be supported by human embryonic fibroblast and human adult fallopian tubal epithelial feeder layers for 20 passages in medium supplemented with

26 15 human serum (79). Using human embryonic fibroblasts (XHEF), a new human ES cell line was then established from a human blastocyst with medium containing 20% human serum. Human ES cells supported by human feeders look and grow like those on MEF feeders. They also maintain normal ES cell morphology, expressed Oct4 and pluripotency associated cell surface markers, along with forming teratomas in severe combined immunodeficiency (SCID) mice. In another study, human foreskin fibroblast were used to grow and maintain human ES cell lines for 70 passages in serum-free medium (80). Similarly these cells retained normal human ES cell morphology and karyotype, expressed SSEA4, TRA-1-60, TRA-1-81, and formed teratomas in SCID mice. Subsequently many new human ES cell lines were successfully established using human foreskin fibroblasts (HFF) as feeder cells with SR (77). Human ES cells established on HFFs showed little difference compared to standard growth on MEFs and maintained all embryonic stem cell like features. A more recent study used XHEF in a defined and xeno-free medium, to successfully support long term cultures of pluripotent human ES cells (22). Despite the removal of FBS, the introduction of SR, and the introduction of human feeder layers, research is active in developing both a xeno- and feeder-free culture system. The disadvantage with feeder layers is that they are highly undefined, producing many unknown proteins and small molecules. Using MEF feeder layers is challenging for large-scale production as they are labour intensive and exhibit lot to lot variability, which adds to the issue of reproducibility and unpredictability when analyzing human ES cells on feeder layers. Moreover, human feeders compared to mouse feeders are also variable, more expensive and difficult to maintain (17) Matrigel/Geltrex Another alternative to MEF feeder layers and co-culturing in general is to culture human ES cells on Matrigel (BD Biosciences) or Geltrex (Invitrogen) coated dishes. Matrigel and Geltrex are both gelatinous extracellular extracts produced from mouse sarcomas that have been shown to be able to support the long-term maintenance of human ES cells. The main components in Matrigel include many ECM proteins such as laminin-111, collagen IV, and entactin, as well as heparan sulfate proteoglycans, and growth factors (81). Some of the growth factors detected in low concentrations include TGF-β, bfgf, epidermal growth factor (EGF), insulin-like growth factor 1, and platelet-derived growth factors (PDGF) (82). Thus, Matrigel is an undefined matrix, which makes interpretation of experiments difficult. Nevertheless, the first cultures supported by Matrigel used MEF-conditioned medium (MEF-CM) (83). Human ES cells under these feeder-

27 16 free conditions expressed Oct4, SSEA-4, Tra-1-81, Tra 1-60, and stained positive for alkaline phosphatase activity. Furthermore they remained undifferentiated for 180 days and retained their differentiation potential as demonstrated with embryoid body and teratoma formation. However, the use of MEF-CM does not alleviate the technical challenge for large-scale production of human ES cells as many MEFs cultures will need to be grown. This also does not remove the variability or the use of animal products in using MEFs. It is possible for Matrigel to support long-term maintenance of human ES cells in serum-free SR medium if a high concentration of bfgf (24 to 36 ng/ml) is supplemented in the medium (84). In SR without supplemented bfgf or with concentrations less than 12 ng/ml, human ES cell cultures lose their typical colony morphology and pluripotency markers. More recently, the International Stem Cell Initiative Consortium undertook a multi-laboratory evaluation of many academic or commercially defined media for their ability to support longterm maintenance of human ES cells on Matrigel (85). Five separate laboratories tested eight promising stem cell culture media in the absence of feeders for cell attachment, survival, and maintenance over five to ten passages. The media chosen were fully defined and serum-free, unlike FBS containing medium. Human ES cells in SR medium with MEF feeder layers were used as a positive control. Of the eight media, only two, both which are commercially available, mtesr1 (StemCell Technologies) and StemPro (Invitrogen) were able to maintain most stem cell lines through ten passages. With the other test media, cultures were terminated because of no initial attachment or from death or uniform differentiation by the fifth passage. It is thought that part of the reason why there appears to be a significant advantage for commercialized media over academic formulation is quality control and manufacturing process that goes into a commercial media. It may be more difficult for academic laboratories to prepare non-commercialized media at the same quality even when key growth factors and specialized additives were purchased centrally and shipped to participating laboratories. More importantly, only mtesr1, StemPro and one other formulation contained agonists for both FGF and activin/nodal/tgf-β pathways, which have been shown to be important in maintaining human ES cells. mtesr1 and StemPro are both defined media, but they are not xeno-free as they contain animal products such as BSA and zebrafish FGF2 (86). Additionally, Matrigel itself is not applicable with clinical therapies because it is derived from mouse sarcomas and is highly undefined, containing over 1800

28 17 proteins (87). As with the use of MEF or MEF-CM, the possibility of exposure to animal pathogens and non-human sialic acid still exists Beginning to Define the Culture System While Matrigel is able to support the long term maintenance of human ES cells without having the labour intensive issue of simultaneously maintaining feeder cells, it still contains undefined animal products and suffers from lot to lot variation. As mentioned, Matrigel presents many ECM proteins to human ES cells such as laminin-111, collagen IV, entactin and heparin sulfate proteoglycan. Both mouse and human feeder layers also lay down laminin, fibronectin, collagen I and IV, while human ES cells themselves have been shown be able to secrete laminin-511 and nidogen 1 (27, 88). If the critical ECM components for human ES cell pluripotency and selfrenewal can be identified, a defined matrix for human ES cell culture can also be designed. Human ES cells express many integrins, which are cell surface receptors that bind to ECM proteins and transmit extracellular signals by activating intracellular signaling pathways (89). They can influence cell proliferation, apoptosis, gene expression, and differentiation. Integrins are heterodimeric transmembrane proteins that form a large extracellular domain with a much smaller cytoplasmic domain. There are 18 mammalian α subunits and 8 β subunits that form 24 distinct integrins with specific and non-redundant functions. Analysis of several human ES cell lines with immunofluorescence and fluorescence-activated cell sorting (FACS) of integrin chains detected many subunits including α1, α2, α5, α6, α11, αv, β1, and β5 chains (27, 90, 91). Consequently, this allows human ES cells to form many integrin heterodimer combinations and bind to all major ECM proteins including collagen, laminin-111, -211, -322, -511, -521, nidogen, fibronectin, and vitronectin. In one of the first studies of a defined culture substrate, Xu et al. utilized a laminin-111 matrix coated on plastic in MEF-CM to successfully maintain several human ES cell lines (H1, H7, H9, H14) over several passages (83). The use of a laminin-coated substrate is based on the knowledge that laminin is expressed early in embryos, Matrigel is rich in laminin, and human ES cells express high levels of the integrin α6β1, specific to laminin. However, cultures on laminin in non-conditioned medium differentiated within 2 passages, suggesting that key factors supporting growth of undifferentiated human ES cells secreted by feeders are missing in media not conditioned with MEFs.

29 Supplementing Growth Factors into ES media In line with using ECM proteins to support human ES cells, Amit et al., tested a feeder-free culture using fibronectin and 15% SR medium supplemented with TGFβ1, LIF, and bfgf (92). While bfgf has been previously shown to help maintain human ES cells in combination with SR and MEFs, LIF does not help to prevent human ES cell differentiation (1). They were able to show that on a fibronectin matrix, human ES cells can be maintained for at least 17 passages if both bfgf and TGFβ1 were supplemented in the media. If TGFβ1 was missing, regardless if LIF was present, cell differentiation was high and complete differentiation occurred after a few passages. Conversely, if bfgf was missing and only TGFβ1 was present, human ES cells remained undifferentiated, but proliferated slowly until no cells remained after 15 passages. In a subsequent report, Beattie et al. demonstrated that activin A, another member of the TGF-β superfamily is able to successfully maintain pluripotency of human ES cells (93). They showed that a feeder-free culture system using laminin-111 and SR medium with supplemented bfgf, activin A, keratinocyte growth factor (KGF), and nicotinamide (NIC) supported human ES cells for over 20 passages. Removal of activin A from the culture resulted in differentiation in one week as shown by loss of Oct4, Nanog, and TRA If activin A was retained, but KGF and NIC were removed, the cells remain undifferentiated, but their growth rates were significantly slowed. Activin A mrna was detected in MEFs and the secreted protein was found in MEF- CM, suggesting it is one of the beneficial factors secreted by MEFs. Another report also supported the role of activin A in maintaining pluripotency and self-renewal of human ES cells (94). Pluripotent human ES cells were maintained in SR medium containing activin A as the only growth factor for over 20 passages. The culture was not grown on defined ECM proteins, but on Matrigel, which contains a number of other growth factors such as TGF-β and bfgf (82). Nevertheless, an inhibitor of activin A, follistatin, was shown to be able to reduce the expression of Oct4 and Nanog in a dose-dependent manner, indicating that activin A was necessary to maintain pluripotency markers (94). Both activin/nodal and TGF-β1 are soluble members of the TGF-β superfamily that activate the same intracellular R-Smad proteins (95). In the TGF-β superfamily, ligand binding results in the heteromeric binding of type I and type II receptors that contain serine/threonine kinase domains. There are five type II receptors that phosphorylate and activate the seven activin receptor-like kinase (ALK) type I receptors, which in turn activate R-Smads that with other factors regulate

30 19 gene expression. There are two subclasses of R-Smads, Smad2 and Smad3; and Smad1, Smad5, and Smad8, which form the two major signaling pathways of the TGF-β super family. Both Activin A and Nodal share the same type I and type II receptors (ALK4; ActRIIB), while TGFβ1 favours the TGFβ1 receptors (ALK5; TβRII), but all three factors activate the Smad2 and Smad3 signaling pathway. Conversely, bone morphogenetic proteins (BMPs) bind to type I receptors that activate Smad1, Smad5, and Smad8. BMP-4 promotes human ES cells towards differentiation into trophoblast cells, but BMP-4 maintains pluripotency in mouse ES cells (96, 97). In the previous study by Beattie et al., if BMP-4 was supplemented in place of activin A, differentiation was quickly induced in human ES cells (93). The TGF-β superfamily regulates many cellular responses depending on species and cell type. TGF-β/activin/nodal signaling plays a large role during embryogenesis and early development as it is important for differentiation including neuroectoderm induction as well as mesoderm and endoderm development (64). However, in undifferentiated human ES cells, the TGFβ/activin/nodal pathway and the downstream Smad2/3 have been shown to be active, while the BMP pathway and the downstream Smad1/5/8 are inactive. When human ES cells are induced to differentiate by removing MEF-CM, Smad2/3 phosphorylation and its subsequent localization to the nucleus is decreased. At the same time, the phosphorylation state of Smad1/5 is observed to be opposite of Smad2/3. Low levels of phosphorylated Smad1/5 are detected in undifferentiated human ES cells, but are increased upon differentiation. The addition of activin A increased Smad2/3 phosphorylation and decreased Smad1/5 phosphorylation, while BMP4 did the opposite. When Smad2/3 phosphorylation was prevented by a synthetic inhibitor, Smad2/3 as well as Oct4 and Nanog levels decreased significantly even in the presence of MEF-CM (64). These results indicate that signaling through the TGFβ/activin/nodal, and specifically Smad2/3 activation is necessary to maintain human ES cells, seemingly contrary to the role of TGF-β in early development. Vallier et al. further demonstrated that the TGFβ/activin/nodal pathway by itself is insufficient to maintain long-term human ES cells in a defined medium without serum (98). Instead, they showed that bfgf in combination with either activin or nodal is capable of long-term maintenance of human ES cells in defined medium without feeder layers or SR. Using a simplified medium with only BSA, lipids, insulin, and transferrin as additives, different growth factors were supplemented individually or in combination to test their ability to support human ES cells.

31 20 While BMP4 and Cripto, a cofactor of Nodal, had no effect, TGFβ1 alone also could not support human ES cells even for a short period of time. In contrast, when activin, nodal, and a high concentration of bfgf were supplemented individually they were able to maintain pluripotency markers for at least one week. Yet by the fifth passage, extensive differentiation was observed and pluripotency markers were lost when only single growth factors were supplemented. The best combination of growth factors observed in this case was activin + bfgf or nodal + bfgf as human ES cells maintained their pluripotency markers for up to 10 passages (98). However, one caveat is that the human ES cells were cultured on undefined substrates as plates were coated with FBS for 24 hours and washed with PBS to remove the serum. The relationship between FGF and activin/nodal signaling was explored using the Alk4/5/7 receptor inhibitor SB (SB) and the FGF receptor inhibitor SU5402 (SU) (99, 100). Human ES cells were cultured in defined medium with either supplemented activin or bfgf. When SU was used to inhibit FGF signaling, the positive effect on pluripotency was lost with bfgf and not with activin. Conversely with SB, the positive effect on pluripotency was lost with not only activin, but also bfgf. As a result, this suggests that while bfgf is necessary for pluripotency, it requires activin/nodal signaling to be active through the Alk4/5/7 receptor pathway (98). Other studies have suggested that the major role of bfgf may be to inhibit BMP4-like activity that is found in unconditioned SR medium (96, 101). In particular, a culture system using Matrigelcoated substrates in SR medium with bfgf and noggin, a BMP antagonist was able to support human ES cells more effectively than bfgf or noggin alone. Levels of a luciferase reporter, responsive for BMP/Smads increased as SR concentration increased, but decreased with higher concentrations of bfgf or noggin, which included a synergistic effect. However, Vallier et al., reported that the mechanism behind FGF s effect on pluripotency does not seem be either promoting Smad2/3 or inhibiting BMP activated Smad1/5/8 nuclear localization (98). The discrepancy may be partly attributed to the use of undefined culture substrates, Matrigel or FBS coated dishes. Alternatively, it is reported that bfgf may maintain undifferentiated human ES cells indirectly by stimulating fibroblast-like cells produced from differentiating human ES cells (hdfs) to secrete supportive growth factors including insulin-like growth factor II (IGF-II) and TGFβ1. Bendall et al., observed that while the FGF receptor 1 (FGFR1) was restricted to the non-colony

32 21 forming, Oct4-negative hdfs, the IGF1 receptor (IGF1R) was only present on human ES cells. Using antibody inhibition of IGF1R resulted in the ES cell death, while inhibition to FGFR1 did not lead to cell reduction, but progressive differentiation. Attempt to rescue IGF1R inhibition with bfgf had no appreciable effect, but adding IGF-II to rescue FGFR1 inhibition resulted in restoration of ES cell number. Furthermore, IGF-II can support human ES cells on Matrigel in the absence of bfgf for over 12 passages in non-conditioned SR medium. Indeed, they also found both IGF-II and TGFβ1 expression in hdfs was increased in human ES cell cultures by adding bfgf. However, Baxter et al., reported that under more defined conditions using a fibronectin substrate and without SR, IGF-II was unable to substitute for bfgf. Human ES cells in these conditions with IGFII became differentiated after 4 passages, but bfgf cultures maintained Oct4 and Nanog. In addition, using immunostaining they were able to show colocalisation of FGFR1 with Nanog positive cells. These results emphasize the variability that is seen across different human ES cell lines as well as between different research labs. More recent studies have shown a direct role for TGFβ/activin/nodal and BMP signaling in regulating pluripotency in human ES cells. Both TGFβ/activin/nodal-activated Smad2/3 and BMP-activated Smad1/5/8 are able to bind to the Nanog promoter (102, 103). In undifferentiated cells, the Nanog promoter is bound by Smad2/3, but bound by Smad1/5/8 in differentiated human ES cells. The activity of the Nanog promoter is reduced by the withdrawal of TGFβ1 and addition of SB, and increased highly with the addition of activin (103). Oddly, while the addition of BMP4 or the removal of bfgf in the medium did not directly reduce Nanog promoter activity, they did significantly increase the reduction of activity initiated by activin/nodal withdrawal or inhibition by SB. Of the three core human ES cell transcription factors, Nanog is nearly lost in all cells within 24 hours of TGFβ/activin/nodal signaling inhibition or absence (102, 103). In contrast, Oct4 expression decreased gradually when treated to the same conditions. The loss of Nanog expression when activin/nodal signaling is inhibited by SB results in the appearance of differentiating cells expressing neuroectoderm markers (102). In contrast, an overexpressing Nanog human ES cell line is able to retain pluripotency markers and does not express any neuroectoderm markers when activin/nodal signaling is inhibited. The overexpression of Oct4 or Sox2 cannot prevent neuroectoderm specification in the absence of activin/nodal signaling. Furthermore it is apparent that Nanog is blocking the neuroectoderm specification that is being induced by FGF signaling (98). Supplementing the FGF inhibitor SU in

33 22 the presence of SB in wild-type human ES cells leads to significantly reduced expression of neuroectoderm markers, indicating the requirement of FGF in neuroectoderm differentiation. Similarly, Nanog repressed ES cells grown with activin and SU expressed very few neuroectoderm markers, but also could not restore pluripotency markers. As a result, FGF signaling is still important in maintaining pluripotency as well as neuroectoderm specification during differentiation. Thus activating the activin/nodal signaling pathway maintains Nanog expression, which prevents FGF-induced neuroectoderm differentiation. In return, Nanog is shown to bind directly to Smad2/3 proteins and limit their transcriptional activity. Past studies in mice and amphibians have implicated high activin/nodal signaling in endoderm differentiation (104). When human ES cells overexpressing Nanog are exposed to culture conditions promoting endoderm differentiation with activin, the appropriate levels of endoderm markers seen with wild type human ES cells are not found (102). Increasing doses of extracellular signals such as activin, BMP or bfgf are not able circumvent the inhibitory effect of Nanog and promote endoderm markers. Evidence shows that Nanog binds directly to Smad2/3 and reduces transcriptional activity, indicating a negative-feedback loop where the maintenance of Nanog through activin/nodal comes back to inhibit its intracellular pathway through Smad2/3. As a result, by maintaining Nanog expression through activin/nodal signaling, both neuroectoderm and endoderm differentiation is inhibited in human ES cells (Figure 2-2). Figure 2-2: Model showing the role of Nanog in maintaining human ES cells. Nanog is upregulated by activin/nodal, which blocks neuroectoderm differentiation by FGF signaling.

34 23 Nanog also inhibits Smad2/3 in order to reduce endoderm specification by activin/nodal signaling. (Source: Vallier et al. (102) ) Research is ongoing to discover new signaling pathways as well as to characterize feeder conditioned media (CM) to discover soluble factors that may regulate signaling pathways promoting self-renewal and pluripotency in hescs ( ). Wnt signaling is another reported pathway that is active in maintaining pluripotency and self-renewal in not only human ES cells but also mouse ES cells (65). The canonical Wnt pathway is initiated when Wnt binds to the Frizzled receptor on the cell membrane and activates downstream signaling that inactivates glycogen synthase kinase-3 (GSK-3). This disrupts the destruction complex that includes a number of proteins including GSK-3 that targets downstream β-catenin to be degraded when Wnt is not present. As a result, β-catenin accumulates and moves into the nucleus to co-activate target genes with T-cell-specific factors (TCFs) (108). Specific GSK-3 inhibitors such as LiCl and more so 6-bromoindirubin-3 -oxime (BIO) have been used to efficiently activate the Wnt signaling pathway (65). Similar to human ES cells grown in MEF-CM, ES cells treated with BIO show nuclear accumulation of β-catenin, whereas cells in non-conditioned medium have no nuclear β-catenin accumulation. More importantly, human ES cells treated with BIO remained undifferentiated and expressed Oct4 and Nanog after 7 days in culture with Matrigel. In contrast, ES cells treated with an inactive analog of BIO differentiate as the expression of Oct4 or Nanog is reduced. Similar undifferentiated ES cells are seen in short term cultures when the Wnt pathway is activated directly with Wnt3a protein. However as mentioned, the TGFβ/activin/nodal signaling pathway through Smad2/3 must be activated to maintain human ES cells. While BIO is able to maintain higher levels of nuclear phosphorylated Smad2/3 than non-conditioned medium, neither BIO nor MEF-CM are able to maintain phosphorylated Smad2/3 when the SB inhibitor is supplemented (64). Consequently, without Smad2/3 activation, the Wnt signaling pathway alone is not able to maintain human ES cells. Moreover, contradictory results were reported in a subsequent study, demonstrating that the Wnt signaling pathway was not sufficient to maintain pluripotent human ES cells (109). Supplementing soluble Wnt3a in basal ES media that included bfgf and insulin was not able to maintain pluripotent human ES cells using Matrigel beyond the first passage. Adding Wnt3a conferred a cell survival/proliferation advantage over basal ES media as more cells were harvested at the end of the first passage compared to absence of Wnt3a. This advantage was

35 24 confirmed using specific Wnt antagonists that blocked Wnt3a binding to its receptors as cell numbers were reduced to the minimum levels displayed in the absence of Wnt3a. Despite the increased cell numbers, the presence of Wnt3a actually reduced the percentage of colony forming units expressing ALP compared to just basal media. Lastly, using a luciferase reporter, evidence showed that β-catenin/tcf transcriptional activity was very low in undifferentiated human ES cells cultured in MEF-CM, but significantly higher in differentiated cells. Supplementing Wnt3a into MEF-CM was able to increase transcriptional activity, but nonetheless it appears that activation of Wnt pathway is not indicative of undifferentiated cells (109) Defined Culture Substrates As more is being understood about how different signaling pathways are able to maintain pluripotency and self-renewal in human ES cells, the more defined different culture systems are becoming. Ludwig et al., derived two new human ES cell lines with defined conditions using ECM proteins and a now commercially available medium known as TeSR1 (now known as TeSR2), which is the xeno-free version of mtesr1 (110). TeSR1 contains a DMEM/F12 base with human serum albumin, vitamins, antioxidants, and lipids along with several growth factors that include TGFβ, bfgf, LiCl, γ-aminobutyric acid (GABA), and pipecolic acid. Removing any of these growth factors reduced the expression of pluripotency markers or proliferation. bfgf was especially important as its absence not only reduced pluripotency, but also greatly reduced cell proliferation (103, 110). Other studies have demonstrated that increasing concentrations of bfgf increases its ability to maintain human ES cells (111, 112). Using Matrigel with TeSR1, several human ES cell lines were shown to be able to grow robustly for months, and exhibit normal karyotype and teratoma formation (110). TeSR1 was able to support the derivation and the long term maintenance of human ES cells using a defined human ECM mix of 10 µg/cm 2 collagen IV, 5 µg/cm 2 laminin, 5 µg/cm 2 fibronectin, and 0.2 µg/cm 2 vitronectin. These lines did not express the non-human sialic acid Neu5Gc, but became karyotypically abnormal after several months in culture, which may because of poor cloning efficiency combined with single cell passaging. When human ES cells are switched to new culture conditions, they may undergo a phase of adaption that might induce karyotype instability and other changes.

36 25 However, a recent study by Hakala et al. could not maintain human ES cells on defined human ECM protein mix and TeSR1 medium described by Ludwig et al. (113). Human ES cells could not be cultured for more than 7 passages until the culture was uniformly differentiated. Additionally, they tested the culture medium described by Amit et al. and Vallier et al., using SR, bfgf and TGFβ1 or bfgf and activin A, respectively (92). Neither medium could sustain undifferentiated human ES cultures in their hands for more than two passages using fibronectin. Instead, several lines of human ES cells were maintained undifferentiated on Matrigel for over 30 passages using mtesr1, suggesting that the undefined xenogeneic substrate is currently superior to purified human ECM proteins. Nevertheless, Li et al. reported a defined and xeno-free culture system for the maintenance of human ES cells using plastic surfaces coated with 2 µg/cm 2 of laminin from human placenta (112). The type of laminin from the placenta is not identified, but it is considered to be laminin- 211/221. A defined medium consisting of X-VIVO 10 base instead of DMEM was used with a very high concentration of bfgf (80 ng/ml) to support human ES cells for up to 40 passages. They also tested the effect of bfgf at 0, 8, 40, and 80 ng/ml in non-conditioned X-VIVO 10 medium on Matrigel coated surfaces. At 0 and 8 ng/ml, differentiation was seen on Matrigel within 2 or 6 passages respectively. With 40 or 80 ng/ml, pluripotency was maintained for over ten passages, but a higher portion of cells exhibited ES cell like morphology with 80 ng/ml. This further underlines the importance of bfgf in human ES cell cultures in maintaining pluripotency and self-renewal. Other studies have investigated the binding of human ES cells to different types of purified laminin proteins (90, 91). Multiple reports suggest that human ES cells are able to synthesize and lay down laminins-111 and -511, which are both critical for early mouse embryogenesis (114). One study showed that human ES cells produce mostly laminin-511/-521 over laminin-111 (91). Moreover, different isoforms of laminin have varying abilities to support human ES cells pluripotency and self-renewal. A defined culture medium StemPro, which contains bfgf and activin A was able to maintain undifferentiated human ES cells on both individually coated laminin-511 and laminin-111 for up to 5 passages (90). Human ES cells on laminin-511 were able to proliferate and self-renewal as robustly as on Matrigel, but on laminin-111, they grew much slower. In addition, although substrates coated with either laminin-511 or -111 were capable of maintaining Oct4 and Sox2 in human ES cells, on laminin-511 they expressed lower levels of

37 26 differentiation markers compared cultures on laminin-111. This was supported by a second study that showed laminin-511 and -111 maintained human ES cell markers after 10 passages in MEF- CM. Another isoform of laminin, -332 has also been reported to maintain the expression of undifferentiated markers, but has been found to promote greater cell proliferation (91). Rodin et al. performed a similar experiment, comparing cell adhesion of human ES cells across different matrices including laminin-511, -332, -411, -111, and Matrigel in defined media (26). Human ES cell adhesion area on laminin-511 was 1.6 times larger than Matrigel and 1.2 times larger than laminin-332. Both laminin-111 and -411 had smaller adherence areas compared to Matrigel. Of the three human ES cell lines tested, all three expressed the laminin chains necessary to synthesize laminin-511, and -111, but not They also showed that using defined media mtesr1 containing BSA and xeno-free TeSR1 containing human serum albumin, human ES cells were able to grow undifferentiated for at least 28 passages on laminin-511 (5 µg/cm 2 ). These cells retained pluripotency markers, normal karyotype, and proper teratomas formation. This culture system with laminin-511 can also support the self-renewal and pluripotency of two ips cell lines in a xeno-free manner. As reported earlier by Amit et al., fibronectin has also been successful as the sole component of a substrate matrix in supporting human ES cells (92). In that study, SR was used in combination with several growth factors, but because SR is animal-derived, a more recent study used N2B27 supplement to substitute SR in a DMEM/F12 base (115). In this novel chemically defined medium, Liu et al. tested a variety of different growth factors including activin A, bfgf, nodal, TGFβ, and Wnt3a to support human ES cells on Matrigel. They found bfgf at high concentrations (100 ng/ml) to support self-renewal better than other growth factor combinations. On defined matrices, N2B27 and bfgf was unable to support human ES cells for more than 7 days with collagen IV or vitronectin. On laminin, human ES cells were maintained for four passages, but fibronectin maintained human ES cells for more than 13 passages in a defined medium. Following this study, Baxter et al. also used fibronectin to support human ES cells for over 10 passages in medium containing the N2B27 supplements and bfgf, activin A, and neurotrophin 4 (NT4) (116). Baxter et al. were not able to support human ES cells on either of the culture systems described by Liu et al. on fibronectin substrates or by Vallier et al. on FBS precoated dishes, further underlining the variability between ES cell lines and research labs. They found that by adding activin A described by Vallier et al. into N2B27 + bfgf as described

38 27 by Liu et al., human ES cells survived for a few additional passages. However once NT4, which has previously been shown to improve human ES cell survival was added into the combined culture medium, they were able to maintain Oct4 and Nanog expressing cells for over 10 passages (117). Over short term cultures, both fibronectin and human placenta laminin substrates maintained undifferentiated human ES cells, but fibronectin had a higher growth rate (116). Fibronectin has also been shown to support human ES cells on more commercialized medium such as mtesr1 (28). Hughes et al. showed varying levels of robustness of a fibronectin and mtesr1 culture system depending on the cell line. While both H9 and CA1 were able to maintain typical morphology and strong expression of pluripotency markers, only H9 exhibited high proliferation rates for over 10 passages. H9 cells formed embryoid bodies expressing markers of all three germ layers and retained normal karyotype after 10 passages. In contrast, CA1 ES cells cultures were unstable and could only be passaged up to 5 passages. Not to be outdone, few more recent studies have suggested that vitronectin could be an effective substrate using defined medium. Several human ES cell lines have been cultured for at least 5 passages on vitronectin using StemPro and over 10 passages using mtesr1 (27, 90). They proliferated at a similar rate to conventional cultures on feeders, as well as maintaining pluripotency markers, morphology, and a normal karyotype (27). These cells also retained their differentiation potential by deriving cell types from each of the three germ layers when induced to differentiate. Short term adhesion assays were also performed, which found attachment on vitronectin to be as effective as on Matrigel (27). In accordance with other studies, attachment to laminin-111 was very weak, while attachment with fibronectin and collagen IV was moderate. However, only vitronectin and Matrigel could promote the self-renewal and pluripotency of human ES cells in a defined medium. Vitronectin coated surfaces were also used to support the maintenance of human ES and IPS cells using a TeSR improved xeno-free media called Essential 8 (E8) (118). TeSR1 and mtesr1 are not truly defined because they still include high concentrations of human serum albumin or BSA respectively. There is significant lot to lot variability in albumin because other lipids and chemicals can bind to albumin. Chen et al., significantly simplified TeSR from 18 components plus DMEM/F12 to 8 components, including DMEM/F12, while maintaining human ES cell culture performance and removing albumin. They show that the removal of BSA results in significant reduction in cell survival and proliferation, which suggests BSA is directly beneficial

39 28 or inhibiting the toxic effects of other components. The latter is true as the subsequent removal of β-mercaptoethanol (BME), restored human ES cell survival and proliferation. Further optimizations led to the inclusion of just eight components including DMEM/F12, insulin, selenium, transferrin, L-ascorbic acid, FGF2 and TGFβ (or Nodal) (118). The survival and cloning efficiency of human ES cells grown on vitronectin in E8 is as efficient as on Matrigel when ROCK (Rho-associated kinase) inhibitor, which has previously shown to improve cloning efficiency is used during single cell passaging (119). Vitronectin-coated substrates with E8 was also able to support ipsc clones for 20 passages, maintaining normal karyotypes, expression of pluripotency markers, and proper teratomas formation. Table 2-1 summarizes the successful human ES culture systems on defined ECM proteins in non-conditioned medium reported in the literature.

40 29 Table 2-1: Summary of successful human ES culture systems on defined ECM proteins in nonconditioned medium. I-3 I-6 H-9 H1 hescs Culture Matrix Fibronectin Fibronectin Fibronectin Laminin (placenta) ECM density or concentration 5 µg/cm 2 5 µg/cm 2 5 µg/cm 2 Culture medium SFM (SR) SFM (SR) SFM (SR) Supplemented growth factors 4 ng/ml bfgf, 0.12 ng/ml TGFβ1 (all) 2 µg/cm 2 SFM (X- VIVO 10) HSF6 Laminin µg/ml SFM (SR) 50 ng/ml activin A, 50 ng/ml KGF, 10 mm NIC H1 H9 H1, H9 H1 HUES1 HES2 HESC-NL3 FES 29, 30 FES 29, 30 MAN1 HUES1 HUES7 HS207 HS420 HS401 H9 CA1 H1, H9 ipsc hecm mix hecm mix Fibronectin Matrigel Vitronectin Vitronectin Vitronectin Laminin-511 Vitronectin Fibronectin Fibronectin Fibronectin Laminin-511 Laminin-511 Laminin-511 Fibronectin Fibronectin Vitronectin Vitronectin 2.4 Matrix Stiffness Col: 10 µg/cm 2 ; vit: 0.2 µg/cm 2 ; fib: 5 µg/cm 2 ; lam: 5 µg/cm 2 TeSR1 TeSR1 50 ug/ml SFM (N2B27) SFM (N2B27) 5 µg/ml 5 µg/ml 5 µg/ml 3.5 µg/ml 4 ug/ml 50 ug/ml 50 ug/ml 50 ug/ml 5 µg/cm 2 5 µg/cm 2 5 µg/cm 2 Not reported Not reported Not reported Not reported mtesr1 mtesr1 mtesr1 StemPro StemPro SFM (N2B27) SFM (N2B27) SFM (N2B27) TeSR1 or mtesr1 mtesr1 mtesr1 E8 E Passages Reference Amit et al. (2004) 80 ng/ml bfgf Li et al. (2005) 20 Beattie et al. (2005) bfgf, TGFβ ng/ml bfgf bfgf, TGFβ bfgf, activin A ng/ml activin A, 4 ng/ml neurotrophin 4, 40 ng/ml FGF bfgf, TGFβ bfgf, TGFβ 10 5 bfgf, TGFβ Not reported 20 Ludwig et al. (2006) Liu et al. (2006) Braam et al. (2008) Vuoristo et al. (2009) Baxter et al. (2009) Rodin et al. (2010) Hughes et al. (2011) Chen et al. (2011) It is now known that mechanical factors play a major role in cell structure and function (120, 121). Most cells exhibit anchorage dependency, that is they are not viable in suspension even if their

41 30 cell adhesion molecules (integrins) are engaged (122). Rather, they need to be adhered to the ECM through anchorage points called focal adhesions, which are made up of primarily integrins (121). Cells pull on their focal adhesions through their actin-myosin cytoskeleton in order to sense the stiffness of the substrate they are adhered to. In turn, cells respond to the resistance through cytoskeleton organization and other cellular process such as changes in gene expression, proliferation and cell migration ( ). In general, cells respond to stiffer substrates by increasing the amount of organized cytoskeleton filaments and the stability of focal adhesions (121). It was recently shown that valve interstitial cells increase in intrinsic stiffness as substrate stiffness increased from 3 to 144 kpa (126). Traditionally, cell cultures are grown on polystyrene plastic dishes, which are millions of times stiffer than most biological tissues. For example, brain tissue has an elastic modulus around kpa, skeletal muscle around 8-17 kpa and precalcified bone around kpa ( ). Normal liver tissues are also very soft around kpa, but a damaged liver with developing fibrosis may be as stiff as 20 kpa (125). It is clear that cell cultures on plastic (elastic modulus ~ 3 GPa) are not in their physiological mechanical environment, and may behave differently as a result. Evidence shows that hepatocytes grown on stiff collagen substrates are proliferative, spread out, and dedifferentiated, while hepatocytes grown on soft collagen substrates that represent normal liver tissues remain quiescent and differentiated (130). Similarly, cardiomyocytes that are cultured on stiff substrates mimicking fibrotic heart tissue lack striated myofibrils, overwork themselves and thus stop beating after a short time. However, when grown on softer substrates that mimic healthy heart tissues (10-15 kpa) cardiomyocytes maintain a constant beat. With substrates softer than the optimal stiffness, cells beat, but generate little force. Stiffness sensitivity is also seen in myoblasts, as they only develop striations when grown on gels with stiffness emulating normal muscle cells (12 kpa) (128). When grown on softer or stiffer substrates, cells do not striate. Myoblasts that are grown on top of glass-attached myoblasts (12 kpa) will form striations, while the myoblasts underneath only assemble loose stress fibers. They also fail to form striations when grown on softer feeder layers such as fibroblasts (2-5 kpa). Stem cell fate and differentiation are also influenced by the matrix elasticity of the culture system. For example, the stiffness of the matrix affects mesenchymal stem cell (MSC) behavior: MSCs grown on soft brain tissue-like substrates differentiated towards neurons, while moderate muscle-like substrates promoted myogenic cells, and the stiffest bone-like substrates led to

42 31 osteogenic cells (129). MSCs seem to commit to the lineage specified by the matrix mechanical properties even in the presence of opposing soluble factors. The sensitivity of matrix elasticity can also be seen in other adult stem cells including skeletal muscle stem cells (MuSCs), which reside on muscle fibers (131). When cultured in vitro on matching 12 kpa hydrogels, MuSCs selfrenewed extensively, which is in contrast to the excessive cell death seen with growing MuSCs on plastic. In addition, less spontaneous differentiation of MuSCs was observed with 12 kpa hydrogels compared to plastic, suggesting better maintenance of MuSC stemness on soft substrates. The maintenance of MuSC stemness was also verified in vivo (131). MuSCs that were cultured and expanded on 12 kpa hydrogels for one week and subsequently transplanted in mice, engrafted at a much higher rate compared to MuSCs that were cultured on plastic. In fact, MuSCs grown and expanded on 12 kpa hydrogels have a similar engraftment rate compared to freshly isolated MuSCs. Recent research has shown that if mescs are cultured on soft polyacrylamide (PA) gels that mimic the stiffness of mescs themselves (~0.6 kpa) instead of conventional stiff polystyrene substrates, mescs remained undifferentiated even in the absence of supporting soluble factors (Figure 2-3) (132). They were able to stay pluripotent on soft 0.6 kpa substrates without exogenous LIF for at least 15 passages. The study examined the effects of culturing mescs on polyacrylamide gels of varying stiffness including 0.6 kpa, 3.5 kpa, and 8 kpa. On 0.6 kpa mesc colonies were round and compact with or without LIF. However on 3.5 kpa and more so on 8 kpa, colonies were more spread out, irregular, and expressed lower levels of Oct3/4 even in the presence of LIF. Colonies on 3.5 kpa and 8 kpa substrates expressed higher cell traction and stiffness, suggesting that the cells were responding mechanically and biologically over a stiffness change of just 6 to 13 fold (132). Moreover, colonies on 8 kpa gels exhibited a similar morphology to that of colonies on polystyrene dishes, which were stiffer by a factor of ~10 6 (>3 GPa). Thus, the study concluded that soft substrates promote the pluripotency marker Oct3/4 in mescs through low traction and low stiffness dependent gene regulation. This demonstrated the importance of matrix stiffness, but whether or not the same effect is found in hescs has yet to be investigated.

43 32 Figure 2-3: Oct4 expression and ALP staining of mouse ES cells with or without LIF on PA gel versus rigid dish. A) High expression of Oct4 and ALP in mouse ES cells on 0.6 kpa gel in LIF+ medium. B) High expression of Oct4 and ALP in mouse ES cells on rigid dish in LIF+ medium, but arrows point to the presence of differentiated colony. C) High expression of Oct4 and ALP in most mouse ES cells on 0.6 kpa gel in LIF- medium. D) Differentiated mouse ES cells on rigid dish in LIF- medium. (Source: Adapted from Chowdhury et al. (132) ) 2.5 Statistical Design and Analysis of Experiment Introduction The use of a statistical modeling method known as design of experiments (DOE) is advantageous for screening many variables or factors in determining an optimal response. Typically, conventional one-factor-at-a-time (OFAT) experiments are used, where one factor is varied at a time while the others are kept fixed (133). Factorial designs are a common technique used in DOE and they are unique compared to OFAT experiments because factorial designs vary multiple factors together (134). Due to this property, factorial designs can not only quantify main effects of individual factors, but also reveal and quantify the interaction effect between two or more factors. One disadvantage with conventional OFAT experiments is that they are unable to detect the interactions between factors because the factors are adjusted one at a time, while other factors remain constant. In OFAT experiments, several replicates are often required in order to calculate an error and produce a statistically significant result. However in factorial design, because

44 33 factors are varied at the same time, combinations of factor levels can provide inherent replication for individual factors when some factors drop out of the design because it is revealed they do not have a significant effect. Additionally, as the number of factors in a factorial design increases, the number of main effects and interaction effects that can be estimated also increases. In these designs, usually the main effects and low-order interaction effects are the most significant. Three-factor and higher high-order interactions by the sparsity of effects principle are often negligible, and can be removed from the design and can be used as a measure of error (134). As a result, factorial designs are more efficient than conventional experiments k Factorial Designs The 2 k factorial designs are a screening tool as the number of design points is reduced. Only two factor levels, low and high are tested for k number of factors. These screening designs assume that the response is linear as only two levels of the factors are tested for their effect on the response. Figure 2-4 shows an example comparison of the experimental design either by conventional methods or with a 2 k factorial design involving an experiment with three factors. With conventional OFAT experiments, one would first test the experimental system with the presence of drug A. Then, drug A would be lowered back to the original low level or absence, and drug B will be tested and so on. A negative control will be included to observe the baseline response of the system, and the whole design will likely be replicated at least two more times to assess statistical error. With the conventional OFAT experiment, the main effects of individual factors can be estimated.

45 34 Figure 2-4: Comparison of experimental design between OFAT experiment and factorial design of an example 3 factor experiment. In the factorial design setup, a 2 3 full factorial design is presented as all 8 combinations of the three factors with two levels are tested. The table shows the coded values, where -1 refers to the low level and +1 refers to the high level of the engineering values. In this case, -1 can mean the drug is not present and +1 for the presence of a drug at some fixed concentration. Here, some experimental runs have more than just one drug present at a time, and this allows the model to estimate the interaction effects of all 3 factors, in addition to the main effects. However, there are no replicates in this factorial design, and it is not a guarantee that a standard error (SE) can be estimated with only 8 runs. If during the analysis, one of the drugs is determined to have no significant effect on the model then it can be removed and used as a replicate. For example, if drug C was not significant, then runs 5 through 8 are essentially replicates of runs 1 through 4 for drugs A and B. From this, the mean square error (MSE) can be estimated in order to calculate the standard error to test the statistical significance of the main effects and interaction effects with t- tests. Alternatively if all factors are significant, then the MSE can be calculated from the sum of square error of the regression model that is generated by the estimated main effects and interactions. The regression model fitted by least squares is as follows: Ŷ = β 0 + β a X a + β b X b + β ab X a X b (eq. 2-1)

46 35 where Ŷ is the predicted response that is determined by several parameters. β 0 is the average response from all the experimental runs; β a is the coefficient of factor A, which is half the main effect of A; and X a is the coded value for factor A. Similarly β b X b is the coefficient multiplied by the coded value of factor B, and β ab is the coefficient of interaction of factor A and B. Using the regression model, the Ŷ response can then be computed for all the experimental runs and the predicted values can be subtracted from the Y actual response in order to determine the residuals. An important assumption is that the residual to Ŷ plot should be random and that no trend in the residuals with changing Ŷ response is observed. For example, if the residuals continue to increase with increasing Ŷ, then there is an issue with the model as the errors are not normal. With the residuals all random, they are then squared and summed to obtain the sum of squared residuals. The MSE is calculated by dividing the sum of squared residuals with the degrees of freedom of the error, provided that there is at least one degree of freedom for the error. A small mean square error is important as it suggests that there is a small difference between the predicted values from the model with the actual response values. Similarly a smaller MSE results in a smaller standard error of effect or coefficient, as shown: Standard error (Effect) = Standard error (β Coefficient) = (eq. 2-2) As shown, the t-test can then be used to evaluate the significance of the factor effects. The larger the effect, and the smaller the standard error, the more significant the factor will be. t-ratio = (eq. 2-3) Furthermore, the MSE is also used in the analysis of variance (ANOVA) of the regression model in order to determine if the model is significant. A significant model means that the model is better at predicting the actual response given the factor levels than just using the simple mean (β 0 ).

47 36 Table 2-2: An example ANOVA of a statistically significant regression model Degree of Sum of Mean F Ratio Freedom Squares Square Model (Treatment) Error (Residuals) P > F * Total Table 2-2 shows an example ANOVA of a statistically significant regression model that was evaluated by determining and comparing the three sum of squares. As mentioned, the sum of squared error is calculated by subtracting the Ŷ predicted from the Y actual response for all experimental runs. To calculate the sum of squared total deviations, subtract the Y average response from the Y response of every experimental run, square each result and add all squared deviations up. Because the sum of squares of the model and the error has to add up to the total sum of squares, the sum of squared treatment deviation can be calculated by subtracting the sum of squared error from the total, as shown in Table 2-2. Alternatively, the sum of squared treatment deviation can be computed by adding up all the individual sum of squares for each factor and interaction included in the model. To calculate the mean squares, the sum of squares is divided by the degree of freedom. The total degree of freedom is the number of runs in the experimental design minus one. In the example, there are eight total runs in the factorial design, and thus the degree of freedom is seven. The number of degrees of freedom for the model is the number of regressors included in the model, which is the number of main effects (β a, β b, etc ) and interaction effects (β ab, β bc, etc ) in the model. In the example 2 3 factorial design, 8 runs are required for the full factorial, but there are 3 main effects, 3 two-factor interactions, and 1 three-order interactions. As a result, in an unsimplified model or in a case where all main effects and interaction effects are significant, there will be 7 regressors, but only 7 total degrees of freedom. With no degrees of freedom left for the error, the MSR cannot be calculated for the ANOVA or factor t-tests. In fact this is the case for all 2 k full factorial designs. For example, the 2 5 full factorial requires 32 runs and has 5 main effects, 10 two-factor interactions, 10 three-factor interactions, 5 four-factor interactions, and 1 five factor interaction, for a total of 31 regressors. Fortunately, most high-order interactions are usually negligible and many main effects and two-order interactions will also be insignificant. In Table 2-2, the number of degrees of freedom in the model indicates that the regression model has

48 37 been simplified to reduce insignificant effects down to four regressors, leaving 3 degrees of freedom left for the error. The F-ratio is calculated by dividing the model mean square with the MSR Calculating Main Effects and Interactions To determine if the main of interaction effects are significant, more than just the standard error of the effect is required, the size or magnitude of the effects is required as well. One advantage of DOE is that each of the effects is estimated by using all the experimental runs in the design. For example in Figure 2-4, all effects are estimated by eight experimental runs, while using conventional designs, the three main effects are only estimated by three runs each Drug B Drug A +1 Figure 2-5: Geometric view of the 2 3 factorial design with hypothetical observations. Figure 2-5 helps demonstrate how all eight runs contribute to estimating the main effect of any factor. For example the main effect of Drug A is the average of all observations where Drug A is +1 minus the average of all observations where Drug A is -1. In this case, the main effect of Drug A is. Thus, on average, the overall effect of increasing Drug A from the coded value -1 to +1 on the response is In essence the main effect of Drug A is the average of all the responses on the right face of the cube minus the average

49 38 response on the left face of the cube. Using equation 2-2 and the MSE from Table 2-2, the standard error of the effect is calculated to be The t-ratio, computed by dividing the effect by the standard error (equation 2-3) is 19.58, making Drug A a highly significant main effect. The main effects of Drug B and C can be shown to also be statistically significant in a similar manner. Solving for two-factor interaction effects is similar to solving main effects. In particular, the interaction effect of Drug A and Drug B would be half the effect of Drug A when Drug B is +1, minus the effect of Drug A when Drug B is -1. In this case, the interaction effect of AB is * +. The t-test for the AB interaction is significant as the test statistic and p value is 6.27 and respectively. The other interaction effects are much smaller, thus not significant, and are removed from the regression model. To illustrate a nonsignificant effect, the interaction effect of BC is = The magnitude of the effect is comparable to the standard error, which is 1.31 and is deemed not significant as it is not certain that the effect is more than just simple error. An easier method to calculating the main and interaction effects is to construct a table of contrasts as shown in Table 2-3. Table 2-3: Table of contrasts for a 2 3 factorial design Main effects Interaction effects Treatment Drug Drug Drug AB AC BC ABC Y combination A B C (1) a b ab c ac bc abc Divisor The table of contrasts is generated by first constructing a table showing all treatment combinations with their corresponding Y response, similarly to Figure 2-4. Then all interaction effects are included by multiplying the signs of the main effects making up the interaction for each treatment combination. Once this is determined, to estimate any effect, sum all the Y

50 39 responses with the corresponding sign indicated in the particular effect s column and divide by the divisor. In this case, the divisor is four, as it is half the number of factorial runs. Using the table of contrasts, the effect of A and AB and respectively, is the same as calculated previously. By calculating all the effects and removing insignificant effects, a prediction expression can be generated using equation 2-1. Ŷ = *A *B (eq. 2-4) *C *A*B The β coefficients used in the prediction formula are half of the effect because the coded values used in the formula for each effect range from -1 to Center Points and Fractional Factorial Designs 2 k factorial designs can be upgraded by adding center points, which are experimental runs that are set with all the factors at the mid-level. This means that the coded values for the center points are 0, as seen with the example in Figure 2-4, which now has been augmented with three center points. Table 2-4: 2 3 factorial design augmented with 3 center points Run Drug A Drug B Drug C The center points act as a form of replication that allows for an estimation of the pure error, which is the error directly attributed to the error from the experiment. Previously it was not known how much of the sum of squared residuals was as a result of the lack of fit from the model or pure error. Because of this, with the center points it is now possible to construct a lack

51 40 of fit test to determine if the residuals are a result of model inadequacy. If the total error or more specifically the lack of fit error is much greater than the pure error arising from the real life experimental variance, then it suggests that the first order model generated by a factorial design is inadequate to fit the data. This would likely to occur if the data is non-linear as a factorial design is a first order model as seen in Figure Figure 2-6: Significant lack of fit using a first order linear model, but not with a second-order quadratic model. Adding center points is an easy method to test for curvature of the data without adding many additional experimental runs. When curvature or lack of fit is detected, a response surface design such as a central composite design is necessary to model the data as more than two levels are used. As a result, in central composite designs additional quadratic parameters are included in the regression model (eq. 1) Blocking and Fractional 2-Level Factorial Designs 2 k full factorial designs are screening methods that help reduce the number of runs to determine important factors and interactions as only two levels of each factor is tested. However, as the number of factors increase, the number of runs required to perform a full factorial design increases exponentially. Five factor and six factor designs require 32 and 64 experimental runs respectively. One way to reduce the number of runs required is to employ blocking techniques, which splits the experiment into halves, quarters or smaller. Each half of the experiment or block can be performed at separate times and the results are combined together. However, there will be confounding effects that become undetermined. Typically, the highest order interaction effect will be the effect that will be confounded with the block effect because main effects and loworder interaction effects are more important. In the 2 3 factorial design, if two half blocks are

52 41 created, the ABC interaction effect would then be confounded with the block. Using the table of contrasts in Table 2-3, to create two blocks with ABC confounded, group the treatment combinations by ABC+ and ABC-. As a result, the ABC+ block would consist of treatments a, b, c, and abc, while the ABC- block would have (1), ab, ac, and bc. In this way, eight runs are assigned to two separate blocks, with the interaction ABC confounded with the block. To further, reduce the number of experimental runs, a fractional factorial design can be used, where only one of the two or more blocks is performed. However, this leads to many confounding effects. For example, performing a half fractional factorial of the 2 3, would be called at fractional factorial and would consist of only four runs. Performing either the ABC+ or ABC- block will lead to the same results and the same confounding effects. To determine the confounding effects, set the interaction (ABC) that is confounding the block effect to I = ABC, which is called the generator or defining relation. Any factor that is multiplied by I is equal to that factor, and any factor squared is equal I. For example (I = ABC) * A = (A = A 2 BC = BC), thus A and BC are confounded, meaning that the effect calculated cannot be distinguished between A and BC. Similarly, B and AC are confounded and so are C and AB. In this way, because main effects are assumed to be dominant over interaction effects, the calculated A = BC effect is attributed to be the main effect over the interaction effect, and thus this design does not have enough resolution to resolve more than the main effects. Similarly, creating two blocks in a 2 5 factorial design will confound the ABCDE interaction with the block effect. Both of the ABCDE+, which is shown in Table 2-5 and ABCDE- blocks will contain 16 experimental runs, making up the total of 32 runs in a full 2 5 factorial design. Table 2-5: ABCDE+ block of a 2 5 factorial design Run Drug A Drug B Drug C Drug D Drug E ABCDE

53 If the experimental runs of both blocks are performed, only the interaction ABCDE will be indeterminable. However if a fractional factorial is used and only one block is performed, the confounding effects can be determined by setting I = ABCDE. With the half fractional factorial, all main effects and two-factor interactions are estimable, only higher order interactions are confounded. Fractional factorial designs are flexible and can be continued to be reduced in experimental size while trading off with the resolution it can resolve. For example, the half fractional factorial can be further reduced to the quarter fractional factorial, where only eight runs are performed. To set up a quarter fractional factorial, four blocks are required to divide 32 runs into groups of eight. As shown in Table 2-6, the division of the treatment combinations is assigned to the signs of ABCD and BCE, which are confounded with blocks. Table 2-6: 2 5 factorial design partitioned into the signs of ABCD and BCE Run Drug A Drug B Drug C Drug D Drug E ABCD BCE

54 Unlike in the previous designs with only two blocks, treatments in the four block design can fall into any combination of ABCD+/- and BCE+/-. Here only eight runs are required to perform a fractional factorial, but it is not recommended unless resources are very costly because no interactions can be resolved. The generators here are I = ABCD = BCE = ADE and thus only main effects are detectable as they are confounded with two-factor interactions.

55 44 Chapter 3 Thesis Hypothesis and Objectives Hypothesis: Biomaterials that mimic the ECM proteins and substrate stiffness of human ES cells supportive feeder layers in defined xeno-free media will enable long-term maintenance of human ES cells under xeno- and feeder-free conditions. Aim 1: To quantify the stiffness of MEF, XHEF, and HFF feeder layers using atomic force microscopy (AFM). Aim 2: To screen for optimal combinations of ECM proteins and stiffness in different media that will promote short term self-renewal and pluripotency of human ES cells on soft polydimethylsiloxane (PDMS) substrates. Aim 3: To validate the optimal combinations of ECM proteins and stiffness determined from screening designs in Aim 2 for long-term self-renewal and pluripotency of human ES cells.

56 45 Chapter 4 Overview of Experimental Designs and Methods 4.1 Measurement of the feeder layer stiffness by AFM In order to determine the range of stiffness to use in the screening experiments, the stiffness of feeder layers were measured using AFM. Several different types of supportive feeder layers were chosen to characterize the stiffness of traditional human ES cell culture systems. The following feeder layers were chosen for their ability to derive and support long-term cultures of human ES cells: four XHEF cell lines (Ea, Eb, Ec, Ed), four XHFF cell lines (Fa, Fb, Fc, Fd), and MEFs. Each of the four XHEF cell lines were derived xeno-free from the connective tissues of different fetuses at the end of the first trimester of pregnancy at the Samuel Lunenfeld Research Institute (SLRI), Mount Sinai Hospital (MSH) as described by Kibschull et al. (22). Two of the XHFF lines, Fa and Fb were xeno-free derived and provided by Dr. D. E. Rancourt (University of Calgary) using their method described by Meng et al. (24). Fc was also derived using the same xeno-free method, but was purchased from Millipore (Billerica, MA, USA), while Fd was a non xeno-free XHFF cell line derived with FBS from ATCC (Burlington, ON, Canada). The MEFs were obtained from the ES Cell Core Facility at the SLRI/MSH. The stiffness of the feeder layers was measured by AFM indentation. To prepare feeder cells for measurement by AFM, they were grown and prepared like feeders to be used as co-cultures with human ES cells. Human feeder layers were grown in xeno-free human fibroblast (XHF) medium containing DMEM (Invitrogen, Carlsbad, CA, USA), 15% human serum (Wisent, Mississauga, Canada), 2 mm GlutaMAX (Invitrogen), 0.1 mm nonessential amino acids (Invitrogen), 100 U/ml penicillin, 0.1 mg/ml streptomycin (Invitrogen), 0.1 mm 2-mercaptoethanol (Invitrogen), and 10 ng/ml bfgf (Peprotech, Rocky Hill, NJ, USA). Feeder cells were grown on 100 mm dishes (Becton Dickinson Company, Franklin Lakes, NJ, USA) at 37 C and 5% CO 2. The medium was changed every 2 days until confluency was reached and the cells were then mitotically inactivated with 10 µg/ml mitomycin C (Sigma, Oakville, ON, Canada) in XHF medium for 2 hours. The medium was then aspirated and the cells were washed with PBS (Invitrogen) three times. The cells were then dissociated with TrypLE Select (Invitrogen), centrifuged at 150xg for 4 minutes, resuspended, and plated at confluency (1.2 million cells) on multiple 35 mm dishes. The cells were kept incubated at 37 C overnight to allow cell attachment

57 46 and spreading. MEF feeder layers were prepared in the same way except MEF medium, while similar to XHF medium in all other components, contained 15% FBS (Thermo Scientific, Waltham, MA, USA) instead of 15% human serum was used. Three different cells were indented with AFM from each 35 mm dishes (5 to 10) per cell line. A commercial AFM (Bioscope Catalyst, Bruker, Santa Barbara, CA) mounted on an inverted optical microscope (Nikon Eclipse-Ti) was used to probe mechanical properties of feeder layer cells in contact mode, following published methods (126). The force-indentation measurements were performed using a spherical tip at three distinct spots (3 cells) on the feeder layer, and the indentation rate was set at 1 Hz with the corresponding velocity of 3 µm/s. The spherical tips were made by attaching a borosilicate glass microsphere (radius: 5-10 μm) onto an AFM cantilever (MLCT-D, Bruker, Santa Barbara, CA) using epoxy glue. The nominal spring constant of the cantilever was 0.03 N/m. A trigger force of 1 nn was consistently applied to the cells resulting in an indentation of less than 500 nm. This helped to avoid erroneously measuring a high Young s modulus due to the underlying substrate as the indentation depth did not exceed 15% of the feeder layer thickness. While no explicit correction was made for the finite sample thickness effects, there was also no observed evidence of depth-dependent stiffening from the underlying hard plastic. The Hertz model was applied to the force curves to estimate the Young's modulus with the Poisson ratio of the feeder layer assumed to be 0.4. The indentation was repeated at the same spot of the feeder layer five times and no significant change in the Young s moduli was seen. Because the Young's modulus determined from the Hertz model is sensitive to the cantilever spring constant, the cantilevers were re-calibrated for each 35 mm dishes by measuring the power spectral density of the thermal noise fluctuation of the unloaded cantilever. The AFM measurements were performed in cell media at room temperature. 4.2 Creating physiologically stiffness relevant polydimethylsiloxane (PDMS) substrates Cell culture substrates with elastic moduli that approximated that of feeder layers (as determined by AFM) were fabricated from PDMS. The substrates were used to culture human ES cells for both the initial screening experiments and for the following validation experiments. Curing PDMS substrates in well plates allowed different ECM proteins to be coated on soft feeder-like substrates. In the screening experiments, PDMS substrates were made in 96-well plates (Sarstedt,

58 47 Newton, NC, USA), while for the validation experiments, larger 24-well plates (Becton Dickinson Company) were used. PDMS substrates were prepared using Sylgard 527 (Dow Corning, Midland, MI, USA), a two-component silicone dielectric gel. These inert non-porous silicone polymers were used to create substrates similar in stiffness to the measured feeder layers. The elastic moduli of the PDMS substrates were tuned by mixing the components A and B of Sylgard 527 in different ratios. The following exponential equation given by Calve and Simon describes the elastic modulus, E, dependence on the ratio of components A to B, as determined with mechanical compression testing of Sylgard 527 PDMS substrates (135) : * ( )+ (eq. 4-1) This regression equation only applies for A/B ratios of 2.06 to 0.115, which result in elastic moduli between 0.5 and 100 kpa, a range containing most soft biological tissues (135) 24- and 96- well plates were coated with 250 µl (~1.3 mm thick) and 50 µl (1.6 mm thick) of PDMS respectively, and they were cured at 80 C overnight. To functionalize the PDMS-coated dishes with ECM proteins, the dishes were first sterilized with 70% ethanol for 10 minutes and left to dry in a biosafety cabinet. PDMS coated well plates were treated with atmospheric plasma using a plasma cleaner (Harrick Plasma, Ithaca, NY, USA) for 20 seconds to temporarily increase hydrophilicity and protein adsorption. Wells were coated either individually or in combination, depending on treatment, with murine laminin-111 (Trevigen, Gaithersburg, MD, USA), bovine fibronectin (Sigma), recombinant human vitronectin (Stemcell Technologies, Vancouver, BC, Canada), or human collagen IV (Sigma). The ECM proteins were diluted in PBS and incubated in the wells at 4 C overnight before being removed and washed with PBS to remove excess proteins. The substrates were sterilized with ultraviolet light for ten minutes and kept hydrated until human ES cells were ready for use. Typically, the substrates were seeded with cells the next day following protein incubation. 4.3 Feeder and feeder-free human ES cell cultures The CA1 line of human ES cells was used to screen and validate optimal culture conditions on ECM protein-coated PDMS substrates. To do this, CA1 cells were passaged from feeder-free standard conditions on Geltrex (Invitrogen) in mtesr1 onto ECM protein-coated PDMS substrates. However, they were first grown on inactivated MEF feeder layers in 6-well plates to

59 48 more easily transition CA1 cells onto feeder-free cultures. On MEF feeder layers, CA1 cells were cultured using SR medium containing KO-DMEM (Invitrogen), 20% KO-SR (Invitrogen), 0.1 mm nonessential amino acids, 100 U/ml penicillin, 0.1 mg/ml streptomycin, 0.1 mm 2- mercaptoethanol, and 10 ng/ml bfgf. SR medium was exchanged every second day until confluency was reached in 5-7 days. Fresh inactivated MEF feeders were seeded on 6-well plates coated with 0.1% gelatin (Millipore) at a cell density of cells/6-well plate in MEF medium. They were kept at 37 C overnight to allow attachment and cell spreading before use. Human ES cultures were enzymatically passaged as single cells by incubating TrypLE Select for 3 minutes at room temperature. Once the cells were dissociated, they were centrifuged with equal volume XHF medium at 150xg for 4 minutes, resuspended with SR medium, and split at a ratio of 1:8-12 (~ cells/6-well plate) onto fresh MEF feeders. As mentioned, to prepare CA1 cells for defined feeder-free cultures on both short-term screening and long-term validation experiments, they were adapted onto Geltrex coated well plates (~15 μg/cm 2 ). For a 6-well plate, 1 ml of Geltrex, diluted at 1:100 in pre-chilled F-12K basal medium (ATCC) was incubated at 37 C for 1 hour. Before use, Geltrex coated plates were let to sit at room temperature for 15 minutes before aspirating the medium and immediately plating the cells. CA1 cells on MEF feeder layers were passaged using TrypLE onto Geltrex coated 6-well plates with mtesr1 (Stemcell Technologies). The feeder-free cultures were exchanged with fresh mtesr1 medium every second day until cultures reached confluence (5-7 days). The cells were then dissociated with TrypLE, centrifuged with equal volume XHF medium, resuspended with mtesr1 and split onto freshly coated Geltrex plates at a ratio of 1:8-12 (~ cells/6-well plate). With the beginning of each passage, the ROCK inhibitor Y (EMD Chemicals, Gibbstown, NJ) and Normocin (Invivogen, San Diego, CA, USA) was added to mtesr1 at a final concentration of 10 µm and 100 µg/ml respectively. After 24 hours, the ROCK inhibitor was removed and only Normocin was added to fresh mtesr Experimental design to screen for pluripotency and selfrenewal of human ES cells To select the ECM proteins to be screened for pluripotency and self-renewal, several proteins shown to be able to maintain long-term human ES cell cultures in the literature were chosen. Many of these proteins have also been shown to be secreted by feeder layers and/or human ES

60 49 cells (27, 88). Here, the ECM proteins, laminin-111, fibronectin, vitronectin, and collagen IV as well as substrate stiffness were selected to determine their effects on pluripotency and selfrenewal of human ES cells. A half fractional factorial was employed to screen for important effects as well as to determine possible interactions between these factors. The two levels chosen for laminin-111, fibronectin, and collagen IV were 0 µg/cm 2 for the low level [-1], and 3 µg/cm 2 for the high level [+1] (Fig. A1). For vitronectin, the low level was set at 0 µg/cm 2 [-1], and 1.5 µg/cm 2 for the high level [+1]. The substrate stiffness of the PDMS ranged from 3 kpa [-1] to 15 kpa [+1]. In our testing, ~3 kpa was the softest the Sylgard 527 PDMS substrates can achieve without the material becoming too sticky and difficult to handle. Moreover, to test whether a second order model was required, three center points were added to the design in order to provide true replication and to generate a lack of fit test. The midlevel [0] of laminin-111, fibronectin, and collagen IV was 1.5 µg/cm 2, while the midlevel of vitronectin was 0.75 µg/cm 2. Finally the midlevel for the substrate stiffness was 9 kpa. Adding the three center points to the experimental design increased the number of runs from 16 to 19 (Fig. A2). The data obtained from the factorial design was analyzed with the following first order regression model fitted by the least-squares method (JMP 10, SAS, NC, USA): Ŷ = β where Ŷ is the predicted response, f is the number of factors included in the model, β 0 is the average response, β i is the coefficient of the main effect, βij is the coefficient of the interaction effect, and the X s are the corresponding coded variables. The two Y responses measured in the screen were ALP staining intensity and the number of Oct4+ cells. Additionally, the screen was performed in two different media, mtesr1 and E8 (Stemcell Technologies) and so four replication of the fractional factorial was required. The model was simplified to avoid model bias or overfitting the data using Mallows Cp. Models that had a large Cp relative to p, where p was the number of parameters in the model, including the intercept were considered biased. Here, stepwise regression was stopped, choosing the first model such that Cp < p and Cp was close to p. The substrates for the 19 experimental runs for the screening design were prepared in 96 well plates. CA1 cells on Geltrex coated plates were passaged onto these substrates at a seeding

61 50 density of 4000 cells per well in a 96-well plate in either mtesr1 or E8. The ROCK inhibitor and Normocin were added at a final concentration of 10 µm and 100 µg/ml respectively for the first 24 hours after passaging from Geltrex. Subsequently, the ROCK inhibitor was removed and mtesr1 with Normocin was freshly exchanged every second day. The cells were grown until confluency was reached in one or more conditions (6-8 days) and the entire screen was then fixed with 4% paraformaldehyde (PFA). Staining with ALP and immunostaining for Oct4 was subsequently performed. 4.5 Validating optimal and suboptimal substrate conditions determined from the screening experiments Once optimal substrate conditions were determined from the screen, they were validated by culturing CA1 cells on 24 well plates over multiple passages in mtesr1 and E8. Multi-passage validation was performed because it may take several passages for human ES cells to fully lose their pluripotency and self-renewal even under poor culture conditions. Thus, some hit conditions from the screen may not maintain human ES cells over long-term cultures, despite initial success in the first passage from the screen. To further validate the screen for its ability to define optimal substrate conditions for potentially new culture systems, optimal substrate conditions were compared against suboptimal conditions as defined by the screen. All conditions tested were also compared against their gel-less counterparts on plastic well plates with the corresponding ECM proteins to observe whether or not soft PDMS substrates improved culture performance. CA1 cells on Geltrex were passaged using TrypLE as single cells at a seeding density of cells per well in 24 well plates containing the selected conditions. The cells were seeded with medium containing both ROCK inhibitor and Normocin, but the ROCK inhibitor was removed after 24 hours. Fresh mtesr1 or E8 with Normocin was exchanged every second day thereafter. Once the cells reached confluency (5-8 days) or if cell growth had slowed after one week, the number of cells on each substrate conditions was counted and passaged onto new substrates. Cell counting for passaging and constructing growth curves was performed using the CASY cell counting technology (CASY Model TTC, Roche Applied Science, Laval, QC, Canada) that utilizes electric current exclusion to count cells. Immunostaining for Oct4 and Nanog, along with ALP staining was performed every few passages to evaluate pluripotency. CA1 cultures on each substrate condition were passaged continuously until the cell culture was lost due to cell death or differentiation.

62 ALP staining and immunostaining For ALP staining, cells were fixed with 4% PFA for ten minutes, and washed with PBS twice. All procedures were then performed according to Vector Red ALP Substrate Kit (Vector Laboratories, Burlington, ON, Canada) and the cells were imaged in water. For immunostaining, cells were first fixed with 4% PFA for 45 minutes, followed by permeabilization with 0.1% Triton X-100 (TX-100) for 30 minutes at room temperature. The cells were then washed 3 times for 10 minutes each with 0.1% Tween20 (Sigma) in PBS and blocked with 1% BSA for 30 minutes. For double staining, a primary labeled Oct4 antibody (Mouse Anti-Human-Oct-4 Antibody, Alexa Fluor 488, Millipore) and an unlabeled Nanog antibody (Rabbit Anti-Human- Nanog Antibody, Cosmo Bio, Carlsbad, CA, USA) were both applied at 1:400 overnight at 4 C. After the primary antibodies had been applied, they were washed once with 0.1% Tween20 in PBS and a labeled secondary antibody for Nanog (Alexa Fluor 594 Donkey Anti-Rabbit, Invitrogen) was applied at 1:400 for 45 minutes. The cells were then washed three times with 0.1% Tween20 in PBS and mounted with Vectashield + DAPI (Vector Laboratories). The cells were photographed with a Retiga-2000R camera (QImaging, Surrey, BC, Canada) using fluorescence (X-Cite 120, Lumen Dynamics Group Inc., Mississauga, ON, Canada) through an inverted microscope (Olympus Canada Inc, Markham, ON, Canada). Fluorescing cells were counted using the ImageJ plugin, ITCN (Image-based Tool for Counting Nuclei).

63 52 Chapter 5 Results 5.1 Feeder layer stiffness measured by AFM The stiffness of feeder layers from several human and mouse cell lines was characterized using AFM with the help of Haijiao Liu from the Simmons Lab (Figure 5-1). The data in each group consist of AFM measurements of several Petri dishes (N = 5-10), averaged from three cells per dish. A normality test originally failed (P < 0.05), so Kruskal-Wallis one way analysis of variance (ANOVA) on ranks was used instead. The differences between median values of the treatment groups were found to be statistically significant (P <0.001). The Dunn s method was used to determine a significant difference between feeder lines Fa vs. Ec, Fa vs. Ea, Fd vs Ec, and Fd vs Ea, as shown in Figure 5-1 (P < 0.05). Combining the cell lines, the mean stiffness for XHEF, XHFF, and MEF were 1.81 kpa, 3.61 kpa, and 1.88 kpa, respectively. There was a significant difference in overall mean stiffness between XHEF and XHFF (P<0.05), but not between XHFF and MEF or XHEF and MEF. The average stiffness of all feeder layers was 2.43 kpa. Figure 5-1: Stiffness of various XHEF, XHFF, and MEF feeder layers. The error bars indicate standard error. Statistical significance of P < 0.05.

64 The significant factors in pluripotency and self-renewal of human ES cell cultures determined from the screening experiments The importance of different types of ECM proteins and substrate stiffness on short-term human ES cell cultures was determined using four fractional factorial experiments. Four separate experiments were performed to measure ALP staining intensity and Oct4 positive (Oct4+) cell count in mtesr1 and E8 medium. Originally, the intention was to measure the percentage of cells that remained Oct4 positive instead of the number of Oct4 positive cells. However, after short-term cultures of only one passage, nearly all substrate conditions showed over 95% Oct4 positive cells and very small differences were observed. Thus, the number of Oct4+ cells for each condition was measured instead. The coded values and the corresponding engineering values are shown in Table A Significant factors of human ES cells grown mtesr1 The estimated regression coefficients for ALP staining and Oct4+ cell count in mtesr1 are summarized in Table 5-1. Among the main effects when measured with ALP intensity, fibronectin (34.0, P < ) and vitronectin (27.8, P = ) were the most significant, while collagen IV (10.9, P = ) was also significant, but had a smaller effect. Each of these ECM proteins had a positive influence to the amount of ALP staining observed. Laminin (8.00, P = ) meanwhile had a small, mostly insignificant positive effect on ALP staining. The main effect of stiffness (-5.22, P = ) was not significant, but remained in the model because of its significance in two interaction effects. Several highly negative interactions worked together to reduce ALP staining including laminin*fibronectin (-27.4, P = ), vitronectin*stiffness (- 14.4, P = ), and fibronectin*vitronectin (-12.9, P = ). Moreover, laminin together with stiffness (15.7, P = ) or collagen IV (12.1, P = ) both exhibited a positive significant interaction on ALP staining. With regards to the regression model for Oct4+ cell count, the highly significant main effects included laminin (806.8, P < ), fibronectin (537.6, P = ), and stiffness (-604.6, P < ). The presence of laminin or fibronectin on average increased the Oct4+ cell count, while higher stiffness on average led to less Oct4+ cells. While the main effect of laminin was positive, the model contained several highly negative significant interaction effects, all working with laminin that reduced the number of Oct4+ cells including collagen IV (-475.8, P = ),

65 54 fibronectin (-442.8, P = ), and stiffness (-234.1, P = ). The other significant interactions in the model all contributed positively to the Oct4+ cell count, including vitronectin*collagen IV (723.5, P = ), fibronectin*stiffness (398.3, P = ), and collagen IV*stiffness (468.6, P = ). Table 5-1: Estimated regression coefficients of simplified fitted model for ALP staining intensity and Oct4+ cell count in mtesr1 medium. ALP staining intensity Oct4+ cell count β ± SE P-value β ± SE P-value Intercept ± 3.83 <0.0001* ± 91.7 <0.0001* Laminin (Ln) 8.00 ± ± 99.9 <0.0001* Fibronectin (Fb) 34.0 ± 4.18 <0.0001* ± * Vitronectin (Vn) 27.7 ± * -29 ± Collagen IV (Co) 10.9 ± * 9.75 ± Stiffness (S) ± ± * Laminin*Fibronectin ± * ± * Laminin*Vitronectin Laminin*Collagen IV 12.1 ± * ± * Laminin*Stiffness 15.7 ± * ± Fibronectin*Vitronectin ± * - - Fibronectin*Collagen IV Fibronectin*Stiffness ± * Vitronectin*Stiffness ± * - - Vitronectin*Collagen IV ± * Collagen IV*Stiffness ± * Prediction expression: Y = *(Ln) *(Fb) *(Vn) (Co) *(Fb)(Ln) *(Ln)(Co) *(Ln)(S) *(Fb)(Vn) *(Vn)(S) Y = *(Ln) *(Fb)+ -29*(Vn) *(Co) *(S) *(Ln)(Fb) *(Ln)(Co) *(Ln)(S) *(Fb)(S) *(Vn)(Co) *(Co)(S) The R 2 of the regression model for the ALP staining intensity was and the P-value for the lack of fit test was The R 2 of the regression model for the Oct4+ cell count was and the P-value for the lack of fit test was

66 Optimal substrate conditions for human ES cells in mtesr1 Determining the optimal substrate conditions in mtesr1 using ALP staining intensity can be observed by analyzing the coefficients of each factor and their interactions in Table 5-1. Each of the main effects for the ECM proteins on average increased ALP staining intensity when included in the model, most strongly by fibronectin and vitronectin. However, the largest interaction was a negative interaction between laminin and fibronectin, suggesting that when present together they reduce ALP staining (Figure 5-2a). Because the main effect of fibronectin was much greater than that of laminin, to maximize Y in the prediction expression fibronectin was set at +1 (high level) and laminin at -1 (low level). With laminin at its low level, collagen IV was set to -1 because the positive laminin*collagen IV interaction effect was slightly larger than the collagen IV main effect. Thus, the loss of the collagen IV s main effect was compensated by the laminin*collagen IV interaction, though very little additional ALP staining was predicted as seen in Figure 5-2b. Another small but significant interaction was the negative interaction between fibronectin and vitronectin. However, because the main effect of both fibronectin and vitronectin was much larger than their negative interaction, to achieve the maximum predicted ALP staining, both vitronectin and fibronectin was set to +1 despite the negative interaction. In Figure 5-2c, this can be seen where the addition of both fibronectin and vitronectin increased the prediction of ALP staining greatly, but not additively of both individual effects. Lastly, for stiffness its main effect is nearly negligible, but it was implicated in two significant interactions, a positive laminin*stiffness interaction and a negative vitronectin*stiffness interaction. Stiffness was set to -1 to maximize ALP staining because laminin and vitronectin were set at -1 and +1 respectively, leading to a positive prediction in their respective interaction as seen in Figure 5-2d-e. As a result, the optimal substrate conditions determined from the ALP staining intensity screen in mtesr1 was a 3 kpa PDMS substrate with 3 µg/cm 2 of fibronectin and 1.5 µg/cm 2 vitronectin.

67 56 Figure 5-2: Response surface maps of ALP staining intensity in mtesr1. Similarly, the optimal substrate conditions for the second model in mtesr1, measured by Oct4+ cell count was determined from the coefficients in the right column of Table 5-1. Here, the main effect of laminin, fibronectin, and stiffness were highly significant, while vitronectin and

68 57 collagen IV were insignificant. On average, increasing the amount of laminin or fibronectin, or lowering the stiffness increased the prediction of Oct4+ cells. In terms of interactions, the largest interaction was a positive vitronectin*collagen IV effect. Because vitronectin and collagen IV had no significant individual effect, from a cost and simplicity standpoint both vitronectin and collagen IV were set at -1. However this became advantageous from a optimization standpoint when stiffness was set at -1, which increased the predicted Oct4+ cell count from not only the large negative stiffness main effect, but also from the large positive collagen IV*stiffness interaction. Figure 5-3a-b shows the increase in the predicted response of vitronectin and collagen IV set at -1 when stiffness was lowered from +1 to -1. Furthermore if laminin was increased to +1 on the basis of its high positive main effect, it indeed benefited from several interactions. Since collagen IV and stiffness were set to their low level, having laminin at +1 counteracted the negative laminin*collagen IV and laminin*stiffness interaction effects, further increasing the predicted response (Figure 5-3b-c). Similar to ALP staining intensity, the laminin*fibronectin interaction had a large significant negative effect. Here the positive effect of laminin was greater than the effect of fibronectin, but the positive main effect of fibronectin was also larger than the negative effect of the laminin*fibronectin interaction, suggesting the inclusion of both laminin and fibronectin for the optimal effect. However, this was not the case because of an additional positive fibronectin*stiffness interaction. Because stiffness was previously set at the low level, fibronectin also had to be set at -1 to work synergistically with stiffness, and the addition of this effect was enough to counter the main effect of fibronectin. Accordingly, Figure 5-3d shows having both fibronectin and laminin at the high level had a lower prediction than having laminin alone without fibronectin. Therefore, the optimal substrate condition as predicted by the screen measured by the number of Oct4+ cells was to have just laminin (3 µg/cm 2 ) on 3 kpa substrates.

69 58 Figure 5-3: Response surface maps of Oct4+ cell count in mtesr Significant factors of human ES cells grown E8 The estimated regression coefficients for ALP staining and Oct4+ cell count in E8 are summarized in Table 5-2. For ALP staining intensity, the only significant main effect was vitronectin (54.1, P = ). At the same time, fibronectin (32.7, P = ) was near significant, while laminin (23.6, P = ) had a smaller insignificant effect. All three of these factors on average contributed positively to ALP staining intensity. Both the main effect of collagen IV and stiffness were nonexistent, but remained in the model due to the presence of several significant interactions. The largest interaction was laminin*collagen IV (50.5, P = ), followed by vitronectin*stiffness (32.2, P = ), which was near significant. All the interaction effects positively affected ALP staining intensity, including the other less significant interactions fibronectin*stiffness (24.4, P = ) and vitronectin*collagen IV (0.1678).

70 59 Regarding model fitting for the Oct4+ cell count, the only main effect that contributed to the number of Oct4+ cells was fibronectin (1311.4, P = ), although increasing stiffness ( , P = ) appeared to have a slight negative effect on cell number. Moreover, the two interactions present in the model were a highly negative laminin*vitronectin ( , P = ) effect, and a near significant fibronectin*stiffness (1510.8, P = ) effect that together increased the Oct4+ cell count. Table 5-2: Estimated regression coefficients of simplified fitted model for ALP staining intensity and Oct4+ cell count in E8 medium. ALP staining intensity Oct4+ cell count β ± SE P-value β ± SE P-value Intercept ± 14.1 <0.0001* ± <0.0001* Laminin 23.6 ± ± Fibronectin 32.7 ± ± Vitronectin 54.1 ± * ± Collagen IV ± Stiffness ± ± Laminin*Fibronectin Laminin*Vitronectin ± * Laminin*Collagen IV 50.5 ± * - - Laminin*Stiffness Fibronectin*Vitronectin Fibronectin*Collagen IV Fibronectin*Stiffness 24.4 ± ± Vitronectin*Stiffness 32.2 ± Vitronectin*Collagen IV 23.1 ± Collagen IV*Stiffness Prediction expression: Y = *(Ln) *(Fb) *(Vn) *(Co) *(S) *(Ln)(Co) *(Fb)(S) *(Vn)(S) *(Vn)(Co) Y = *(Ln) *(Fb) *(Vn) *(S) *(Ln)(Vn) *(Fb)(S) The R 2 of the regression model for the ALP staining intensity was and the P-value for the lack of fit test was The R 2 of the regression model for the Oct4+ cell count was and the P-value for the lack of fit test was

71 Optimal substrate conditions for human ES cells in E8 The optimization of the model determined by ALP staining intensity in human ES cultures grown in E8 was observed by analyzing the coefficients of each factor and their interactions in Table 5-2. Similar to previous models, all the ECM proteins generally had a positive effect on the predicted response. Here, vitronectin had the largest and most significant effect, followed by fibronectin and laminin. Collagen IV and stiffness had negative, but negligible effects. Moreover, since all of the interaction effects were positive, all the interacting factors involved worked together to increase ALP staining intensity. As a result, combined with all significant positive main effects, the most optimal substrate condition predicted by this model was all five factors at the high level. Figure 5-4 shows the response surface maps of the two factors with the largest main effect, vitronectin and fibronectin with the other factors held at their mid-levels and then at their high levels. Figure 5-4: Response surface maps of ALP staining intensity model in E8 For determining the optimal conditions in E8 using Oct4+ cell count, the coefficients in the right column of Table 5-2 were used. In this case there were only a few significant or near significant effects. Interestingly, while the positive main effect of fibronectin was near significant, it had little influence on the predicted optimal response because it was counteracted by two other effects. If fibronectin was set to +1, then because of the positive fibronectin*stiffness interaction, stiffness would also have to be set at +1 for a positive prediction from this interaction term.

72 61 However, by using stiffness at its +1, its negative main effect would reduce the predicted response by approximately the same amount as the main effect of fibronectin. In the same way, if fibronectin was reduced to -1 and stiffness was set to -1 to satisfy the interaction term, the predicted Oct4+ cell count would be the same had if fibronectin and stiffness were at +1. Figure 5-5a shows this relationship between fibronectin and stiffness, which did not change with other factors as fibronectin or stiffness had no other interactions. The prediction with fibronectin and stiffness at +1 was slightly higher than -1 because of the slightly greater magnitude of fibronectin s main effect compared to stiffness. However, this difference was negligible, and from a cost and simplicity perspective fibronectin and stiffness were set at -1. The other significant interaction dictating the model is the very large negative laminin*vitronectin effect. As with the other negative interactions, the presence of both factors at the high level or low level will result in the negative effect as seen in Figure 5-5b. Here the main effect of laminin was slightly greater than vitronectin, and thus laminin was set at +1 and vitronectin was kept at -1. As a result, the optimal condition predicted by the Oct4+ cell count model in E8 was laminin at the high level and all the other factors at the low level. Figure 5-5: Response surface maps of Oct4+ cell count model in E Suboptimal conditions defined by screens As part of validating the models derived from the screening experiments, suboptimal conditions defined by the model were also determined. In general the least optimal conditions predicted by the models contained no ECM proteins or just collagen IV on 15 kpa substrates. For example,

73 62 using the ALP staining intensity mtesr1 model, it predicted having either collagen IV or no ECM protein on 15 kpa substrates as the least optimal conditions. This was seen by setting highly favourable main effects fibronectin and vitronectin to -1 (Table 5-1). Additionally laminin was also set to -1 in order to antagonize fibronectin through the highly negative laminin*fibronectin interaction. Figure 5-6a shows the relationship between collagen IV and stiffness when the other three factors were set to -1. As seen in the response surface map, the suboptimal conditions predicted by this model were either coating 15 kpa substrates with 3 µg/cm 2 of collagen IV or with no ECM proteins at all. For the Oct4+ cell count mtesr1 model, it predicted having no protein, collagen IV, or vitronectin on 15 kpa substrates as having zero Oct4+ cells. Similarly, this was observed by setting the highly significant main effects, laminin and fibronectin at -1 and stiffness at +1. The combination of laminin, fibronectin, and stiffness at their unfavourable levels led to a very low prediction of Oct4+ cells, as seen in the response surface map of vitronectin and collagen IV (Figure 5-6b). Only having both collagen IV and vitronectin at +1 led to a moderate predicted response. Otherwise, having no protein or just collagen IV or vitronectin individually on 15 kpa substrates led to very poor predicted Oct4+ cell count. The suboptimal conditions for the two models in E8 were similar to the ones in mtesr1. For the ALP staining intensity model, the main effect of vitronectin, fibronectin, and to a lesser extent laminin were all highly positive towards ALP staining (Table 5-2). Assigning these factors to -1 greatly reduced the predicted ALP staining, while putting the laminin*collagen IV and to a lesser extent the vitronectin*stiffness and fibronectin*stiffness positive interactions into play. With laminin, vitronectin, and fibronectin set at -1, it allowed the predicted ALP staining intensity to be further reduced through these interactions when collagen IV and stiffness were set to +1. Figure 5-6c shows the least optimal substrate condition when collagen IV was coated onto 15 kpa substrates. Similarly having no ECM proteins on 15 kpa substrates had a near zero ALP staining prediction. Lastly, the Oct4+ cell count E8 model predicted having no protein on 15 kpa substrates as the least optimal condition. Fibronectin was the only protein with a main effect that was near significant in this model, and for determining the least optimal conditions it was set to -1. Similarly, stiffness had a negative near significant main effect, and it was set at its high level to

74 63 reduce the predicted Oct4+ cell count. From here laminin and vitronectin can either be at -1 or +1 to produce the highly negative effect of the laminin*vitronectin interaction. Figure 5-6d shows the interaction of laminin*vitronectin and that if they were both set at -1, as in no ECM protein coating on 15 kpa substrates, then the predicted Oct4+ cell count would be below zero. Interestingly, with all else constant, having laminin and vitronectin at +1 would also result in a low-moderate predicted response. Collagen IV was dropped in this model, but if the stepwise regression was stopped before collagen IV was removed, then the least optimal substrate condition would have had only collagen IV on 15 kpa substrates. Figure 5-6: Response surface maps of demonstrating less optimal substrate conditions determined from different models.

75 Validating optimal and suboptimal substrate conditions determined from the screening experiments Three optimal substrate conditions and one suboptimal substrate condition were tested for longterm cultures on 24-well plates. The optimal substrate conditions tested were derived mostly from the mtesr1 models, summarized in Table 5-3. Substrate condition 1 was the optimal substrate condition determined from the ALP staining intensity mtesr1 model. It comprised of fibronectin and vitronectin each coated at 3 µg/cm 2 on 3 kpa substrates. Substrate condition 2 was the optimal substrate condition defined by the Oct4+ cell count model, comprised of just laminin at 3 µg/cm 2 on 3 kpa substrates. This was also the optimal condition determined by the Oct4+ cell count in E8 medium. For substrate condition 3, fibronectin coated at 3 µg/cm 2 on 3 kpa substrates was tested because in both the ALP staining intensity and Oct4 cell count model in mtesr1 there was a highly significant laminin*fibronectin interaction along with a very high fibronectin main effect. Lastly, substrate 4 was the suboptimal substrate condition tested in longterm cultures, containing collagen IV at 3 µg/cm 2 on 15 kpa substrates. Table 5-3: Summary of substrate conditions Substrate condition no. Components 1 Fibronectin 3 µg/cm 2 Vitronectin 1.5 µg/cm 2 Substrate stiffness 3 kpa 2 Laminin 3 µg/cm 2 Substrate stiffness 3 kpa 3 Fibronectin 3 µg/cm 2 Substrate stiffness 3 kpa 4 Collagen IV 3 µg/cm 2 Substrate stiffness 15 kpa

76 65 Figure 5-7: Validation results of different substrate conditions either in mtesr1 or E8. The cultures that were terminated because of poor attachment and survival are marked with a X. Some substrates were tested multiple times, and they are denoted by a second or third darker shaded bar. The arrows indicate the cultures are currently still being maintained CA1 cultures on optimal substrate conditions grown in mtesr1 CA1 cultures were initiated on ECM protein conditions (Table 5-3) prepared on specified PDMS substrates or on regular plastic well plates in both mtesr1 and E8. Figure 5-7 shows a summary of the fate of human ES cell cultures grown in various conditions. Most substrate conditions in mtesr1 were unable to support long-term maintenance of human ES cell cultures past 5 passages. Even with early passages, cultures in mtesr1 grew slowly due to poor initial

77 66 attachment and/or excessive cell death regardless of substrate conditions. CA1 cultures in mtesr1 took 7-9 days to become somewhat confluent enough to be passaged onto new substrates. In contrast, cultures grown in the same conditions in E8 media became completely confluent in 5-6 days. The speed and the number of cells at the end of each passage when grown in E8 were much greater than the same conditions in mtesr1 as seen in the growth curves in Figure 5-8 and 5-9. Moreover, slightly better culture performance in terms of cell numbers and number of passages was seen for substrate conditions on plastic compared to 3 kpa substrates when using mtesr1. Representative phase contrast images of cultures grown in mtesr1 shown in Figure 5-12 also suggest better culture performance for substrate conditions on plastic. While the first passage showed similar amounts of cells on plastic and 3 kpa after five days, by the second passage cells on 3 kpa substrates grew slowly. Other substrate conditions similarly did not grow robustly on 3 kpa after the first passage. However, as mentioned, all cultures in mtesr1 even on plastic were lost as fewer cells remained with each passage until there were no cells remaining. These substrate conditions were unable to sustain continuous ES cell self-renewal in mtesr1, but after the second passage all cultures still retained high expression of Oct4 and Nanog pluripotency markers as seen in Figure The data does not suggest any performance differences in any of the substrate conditions in terms of percent of cells expressing Oct4 and Nanog. Only condition 2 on plastic in mtesr1 survived to passage 5 and reassessment of pluripotency markers showed most cells were still expressing Oct4 and Nanog. ALP staining was also positive with all the substrate conditions grown with mtesr1 (Figure 5-13).

78 Cell number (x10 4 ) Cell number (x10 4 ) C1 3 kpa C2 3 kpa C3 3 kpa C1 plastic C2 plastic C3 plastic Passage Figure 5-8: Growth curves of CA1 grown on various substrate conditions in E8. Cells were grown for 5-6 days before being counted and passaged onto new substrates. Each culture and passage was started with cells C1 3 kpa C2 3 kpa C3 3 kpa C1 plastic C2 plastic C3 plastic Passage Figure 5-9: Growth curves of CA1 grown on various substrate conditions in mtesr1. Cells were grown for 8 days before being counted and passaged onto new substrates. Each culture and passage was started with cells.

79 Percentage of cells expressing plurioptency marker Percentage of cells expressing plurioptency marker % 80% 60% 40% 20% C1, 3 kpa C2, 3 kpa C3, 3 kpa C1, plastic C2, plastic C3, plastic 0% Passage 2 Passage 5 Passage 2 Passage 5 Oct4 Nanog Passage and pluripotency marker Figure 5-10: Percentage of cells expressing Oct4 or Nanog pluripotency marker in CA1 cultures on different substrate conditions in mtesr1. Most substrate conditions in mtesr1 did not survive to passage 5 to access their pluripotency. Oct4 and Nanog percent expression was determined through immunostaining and image analysis. 100% 80% 60% 40% 20% C1, 3 kpa C2, 3 kpa C3, 3 kpa C1, plastic C2, plastic C3, plastic 0% Passage 2 Passage 5 Passage 7 Passage 2 Passage 5 Passage 7 Oct4 Nanog Passage and pluripotency marker Figure 5-11: Percentage of cells expressing of Oct4 or Nanog pluripotency marker in CA1 cultures on different substrate conditions in E8. Substrate condition 2 on plastic did not survive

80 69 up to passage 5 to have its pluripotency markers reassessed. Oct4 and Nanog percent expression was determined through immunostaining and image analysis. Figure 5-12: Representative phase contrast images of CA1 cells on substrate condition 1 in mtesr1 after five days for passage 1 and 2. On plastic, cell numbers were comparable between passage 1 and 2, but on 3 kpa substrates, cell numbers were diminished in passage 2. Scale bars, 100 µm.

81 70 Figure 5-13: Representative phase contrast images of CA1 cells on different substrates in mtesr1 stained with ALP. Scale bar, 100 µm CA1 cultures on optimal substrate conditions grown in E8 While substrate conditions in mtesr1 were unable to maintain CA1 cultures beyond five passages, substrate conditions using E8 were readily able to support continuous growth of CA1 cultures beyond ten passages (Figure 5-7). Only substrate condition 2 on plastic in E8 was unable to maintain CA1 cultures beyond the fifth passage due to poor initial attachment and survival after each passage. When second and third cultures were reinitiated using condition 2 on plastic, very poor initial attachment and cell death led to termination of culture at passage 1. In contrast, substrate condition 2 using 3 kpa substrates was able to maintain CA1 cultures beyond 10 passages. Representative phase contrast images of these substrate conditions grown in E8 are shown in Figure The photo of each substrate conditions in passage 6 was taken after 5 days, similar to the cultures grown in mtesr1 in Figure Comparing the two sets of photos suggest that the growth and/or survival of CA1 cells were better in E8 than in mtesr1. In addition, Figure 5-14 shows that in E8, the substrate conditions on soft substrates had greater growth and/or survival compared to the same substrate conditions on plastic. These observations were supported by the growth curves of the substrate conditions grown in mtesr1 and E8 in Figure 5-8 and 5-9. As mentioned previously, cell numbers at the end of each passage were greater in E8 cultures than in mtesr1 cultures. However, the E8 growth curves also show that the cell numbers for each substrate condition on 3 kpa substrates were consistently higher than the same

82 71 substrate conditions on plastic. Yet, there does not seem to be a clear performance difference among the substrate conditions on 3 kpa substrates in E8. Immunostaining and image analysis of Oct4 and Nanog were performed at passages 2, 5, and 7 for cultures that had maintained growth to those time points. Figure 5-11 presents the data as the percentage of cells expressing Oct4 and Nanog. The results show that all surviving cultures as of passage 7 were still highly expressing two key pluripotency markers, which agreed with morphological observations. In terms of maintaining Oct4 and Nanog, there did not seem to be a performance difference between any of the substrate conditions whether on 3 kpa or plastic in E8 media. All surviving cultures in passages 2, 5, and 7 also stained strongly for ALP.

83 72 Figure 5-14: Representative phase contrast images of CA1 cells on various substrate conditions in E8 after five days. While CA1 cells overall grew faster than cultures in mtesr1, substrate conditions on soft substrates had even more robust growth. Scale bars, 100 µm.

84 73 Figure 5-15: Representative phase contrast images of CA1 cells on different substrates in E8 stained with ALP. Scale bar, 100 µm CA1 cultures on the suboptimal substrate condition CA1 cultures grown on the suboptimal substrate condition (no. 4) were unable to maintain human ES cells past 3 passages either in mtesr1 or E8 (Figure 5-7). When cultured on 15 kpa substrates, as defined by the screening models, CA1 cells initially attached to the substrate, but subsequently detached within 3-5 days. As a result, cultures on substrate condition 4 on 15 kpa substrates in both mtesr1 and E8 were terminated at the first passage. In contrast, when CA1 cells were cultured using condition 4 coated on plastic, they were able to survive up to the third and second passage in mtesr1 and E8 respectively before they were lost. Figure 5-16 shows the cultures on condition 4 on plastic at passage 3 in mtesr1 detaching over the course of several days and eventually terminated.

85 74 Figure 5-16: Phase contrast images of CA1 cells on the suboptimal substrate condition in mtesr1 on plastic. While CA1 cells were able to be maintained to the third passage, they began to detach after day 4 and the culture was lost. Scale bars, 100 µm.

86 75 Chapter 6 Discussion For clinical and therapeutic applications of human ES cells to be a reality in the future, human ES cells and their derivatives must be generated in defined conditions that are free of xenogeneic contaminants. The use animal products increases the risk of transferring nonhuman pathogens as well as substances like nonhuman sialic acid Neu5Gc that may induce an immune response in patients (20). Hence, avoiding all nonhuman animal products is typically required to comply with strict regulatory rules. It is also of interest for non-clinical based research to use defined materials to avoid batch-to-batch variation. It would be difficult to interpret signaling pathways or effects of growth factors if unknown components are prone to variability and are present in culture with human ES cells (17). Several studies have investigated the effectiveness of different types ECM proteins as a defined feeder-free system to maintain human ES cells (Table 2-1). Varying degrees of success have been reported with ECM proteins using a wide variety of different media and cell lines. However, there are some inconsistencies in the literature and no agreement to which ECM protein offers the best culture performance. For example, Braam et al. reported that with defined medium (mtesr1) only purified vitronectin, and not fibronectin, laminin, or collagen IV were able to maintain human ES cells (27). In contrast, Rodin et al. and Li et al. showed that different types of laminin were supportive of human ES cells in defined media, while Vuoristo et al. found that proliferation was greater on laminin-511 compared to vitronectin (26, 90, 112). Adding to this, Liu et al. reported that collagen IV and vitronectin were unable to culture human ES cells for more than 7 days, and on laminin, no more than 4 passages (136). Only on fibronectin could they robustly maintain human ES cells for more than 13 passages with defined medium. More strikingly, Ludwig et al. was able to derive new human ES cell lines using a defined human ECM protein mix containing laminin, fibronectin, vitronectin, and collagen IV in TeSR1 medium (110). Hakala et al., repeating the experiment was not able to support existing human ES cells using the same ECM protein mix and medium, as differentiation occurred beyond the early passages (113). However, they both found that a mix of all four ECM proteins was superior to just individual ECM proteins. Some of the inconsistencies can be explained by the use of different media and human ES cell lines, which suggest that current culture systems are not robust enough. Most previous studies have only tested the culturing performance of each ECM protein

87 76 one at a time on plastic. Varying multiple ECM proteins to estimate the effect of different combinations of ECM proteins has not been studied previously, only as all four proteins in a general mix. In this proof of concept study, we used factorial designs to identify optimal substrate conditions from five factors including laminin, fibronectin, vitronectin, collagen IV, and substrate stiffness. In the screen we also incorporated substrate stiffness as evidence from other stem cell types suggests that stiffness may be important in regulating stem cell fate (129, 132). The screening experiments evaluated the pluripotency and self-renewal of short-term cultures by measuring the amount of ALP staining intensity and the number of Oct4+ cells in both mtesr1 and E8 media. The factorial designs tested the culture performance of many combinations of all five factors. Several different optimal substrate conditions were defined by main effects and interactions determined from four separate screens. Firstly, the data from the ALP staining intensity in mtesr1 demonstrated fibronectin, vitronectin, and low stiffness (3 kpa) was highly associated in promoting ALP staining. Secondly, measuring the number of Oct4+ cells showed that laminin and low stiffness (3 kpa) was highly linked to promoting pluripotency and self-renewal. Common to both sets of data, fibronectin was also highly associated with pluripotency and selfrenewal, but was highly antagonized by the presence of laminin. Thus fibronectin on 3 kpa substrates was also selected for validation. These screens were performed a second time using the xeno-free E8 medium. The results from the ALP staining intensity in E8 were not consistent with ALP staining intensity in mtesr1. Here, the presence of all four ECM proteins on a 15 kpa gel was greatly associated with ALP staining. In contrast, the results using Oct4+ cell count in E8 model determined the same substrate optimal as the Oct4+ cell count in mtesr1, where laminin coated on 3 kpa substrates were most highly associated with pluripotency and selfrenewal. Because these screening experiments were performed for only the first one passage, many different hits or optimums will be generated for short term cultures. Even on unfavourable conditions, it may take more than one passage for cells to lose their pluripotency markers or to stop proliferating. For example, it was initially thought that soluble Wnt3a, which activates the Wnt pathway, was sufficient to maintain pluripotency and self-renewal in human ES cells (65). However, while the activation of the Wnt pathway may increase survival/proliferation, it was shown that beyond the first passage, Wnt activation was not sufficient to maintain pluripotency

88 77 and self-renewal of human ES cells (109). This discrepancy was likely not due to difference in tissue-culture methods or cell line used as both studies used the same human ES cell line (H1), but that the original study only cultured human ES cells for 7 days. Similarly, we had intended to use the percent of cells expressing Oct4 at the end of the screen in order to obtain a measurement of pluripotency instead of both pluripotency and self-renewal. It turns out that almost every condition had over 95% of cells expressing Oct4 at the end of the first passage, and the difference of the results between substrate conditions was not large enough to obtain statistically significant models. In a sense, our screens are sensitive, ensuring that most conditions that have any potential to support human ES cells in the long-term will be identified as hit. As a result, these optimal substrate conditions must be further validated for their ability to support pluripotency and self-renewal in long-term cultures. The selected optimums and suboptimal conditions to be validated are shown in Table 5-3. They were tested in mtesr1 as well as E8 for their ability to support human ES cells in xeno-free media. The ECM components of each substrate condition were also tested on plastic well plates to observe if PDMS substrates and substrate stiffness contribute to culture performance. Interestingly, the data from the validation studies demonstrated that even though the three hit conditions were selected from the mtesr1 screens (one condition was consistent with E8 screen), only E8 was able to maintain most substrate conditions beyond 10 passages (Figure 5-8). It was somewhat surprising that none of the optimal substrate conditions were able to maintain human ES cells in mtesr1 past even 5 passages. While Oct4 and Nanog pluripotency markers were retained, proliferation and self-renewal were not maintained at a level sufficient for longterm cultures (Figure 5-11). Studies by Beattie et al. and Li et al. have shown laminin as a capable ECM protein to support human ES cells beyond 10 passages using similar media to mtesr1 with respect to growth factors (93, 112). Likewise among other studies, Braam et al. and Baxter et al. showed that vitronectin and fibronectin respectively were supportive for 10 or more passages in mtesr1 or similar media (27, 116). Yet our substrate conditions of fibronectin + vitronectin (condition 1), laminin (condition 2), and fibronectin (condition 3) were not able to maintain CA1 human ES cells in mtesr1 past 5 passages. Part of the inconsistency may be due to differences in cell lines, media, or tissue-culture techniques. For example, Hughes et al. grew both H9 and CA1 human ES cell lines on fibronectin-coated plates in mtesr1 for 10 and 5 passages respectively

89 78 (28). While both cell lines retained pluripotency markers, cell growth with CA1 was not as robust as H9 human ES cells, making it hard to maintain a stable culture, free of differentiation beyond 5 passages, which was consistent with our results. Furthermore in our cultures with mtesr1, we did not observe any improvement in pluripotency or self-renewal with culturing CA1 cells on soft PDMS substrates instead of on plastic. Instead, survival and proliferation of human ES cells do appear to be slightly better on plastic (Figure 5-10). This is in general agreement with a recent study demonstrating that the proliferation/survival of human ES and ipscs were not affected by ECM stiffness (137). Keung et al. cultured human ES cells in serum-free medium containing bfgf and TGF-β and ipscs in mtesr1 on Matrigel-coated polyacrylamide gels of 0.1, 0.7 and 75 kpa stiffness. After a short time period of 3 days, the overall cell density was measured and no difference was seen across the range of stiffness tested for either cell type, despite cell spreading being several times greater on stiffer gels. Our data with mtesr1 also seem to agree that selfrenewal is not greatly affected by substrate stiffness. In contrast, using xeno-free E8 medium with the optimal substrates, we were able to maintain CA1 cells beyond ten passages. All three substrate conditions on PDMS substrates or plastic, except condition 2 on plastic, maintained pluripotency markers and had robust self-renewal beyond 10 passages. Cultures in E8 took only 5-6 days to become confluent and have cell counts often between cells per well. Meanwhile, cultures in mtesr1 were grown for 8 days, but still rarely had cell counts over cells per well and eventually stopped growing. The difference in culture performance was impressive and unexpected as E8 was a fairly new formulation that had not been readily used in the literature thus far. E8 is an improved and simplified TeSR media that does not include highly variable human/bovine serum albumin or other components such as BME. E8 has been shown to support long-term undifferentiated proliferation of human ES cells comparably to TeSR on Matrigel, as well as in initial testing with vitronectin (118). However, our substrate conditions showed significant improvement in proliferation and self-renewal in E8 compared to mtesr1. Whether the higher cell counts were a result of greater proliferation, survival, or initial attachment remains to be seen. The difference in ability to maintain pluripotency markers between E8 and mtesr1 was negligible, but long-term performance could not be compared as cultures on mtesr1 did not survive long enough. The positive effect of soft 3 kpa substrates on culture performance that was predicted by the screening experiments was evident in E8. This was most apparent with substrate condition 2,

90 79 which containing just laminin had difficulties maintaining a stable culture on plastic. While the cells still expressed pluripotency markers and morphologically looked like ES cells, initial attachment and/or self-renewal was very poor. When we attempted to reinitiate the culture on this condition from Geltrex, both attempts failed because of poor initial attachment. Yet, when condition 2 was coated on 3 kpa substrates, pluripotent human ES cells proliferated and selfrenewed robustly for over 10 passages (Figure 5-9). In addition, when we compare all substrate conditions on 3 kpa with plastic in E8, we find that 3 kpa substrates had greater proliferation/survival than on plastic across all conditions. This was consistent across all passages suggesting that unlike our cultures in mtesr1 and work by Keung et al., substrate stiffness does affect human ES cell self-renewal when grown in E8. In terms of pluripotency, both Oct4 and Nanog were expressed by most cells regardless of substrate stiffness. At least up to passage 7, there were no consistent differences in Oct4 and Nanog expression of CA1 cells whether grown on 3 kpa substrates or on plastic in any ECM protein combination. In order to observe differences in maintaining pluripotency markers among different substrate conditions, longer cell passages and different human ES cell lines should be tested. Chowdhury et al. demonstrated that using soft PA gels that matched the intrinsic stiffness of mouse ES cells improved self-renewal and pluripotency over plastic dishes (132). In fact using soft substrates, they were able to maintain pluripotent mouse ES cells without supplemented growth factors for 15 passages. Thus we may see greater differences in culture performance among different substrate conditions if lower concentrations of growth factors are used in mtesr1 and E8. The self-renewal and pluripotency of MuSCs was also improved by culturing them on substrates with stiffness that was matched to that of the stem cells (MuSCs) (131). When grown on these substrates, MuSCs did not undergo excessive cell death seen on plastic substrates, and the overall cell number was increased. Less spontaneous differentiation of MuSCs was also observed with soft gels compared to plastic, suggesting better maintenance of MuSC stemness on soft substrates. Here, we similarly used soft substrates aimed at mimicking the substrate stiffness of the microenvironment of human ES cells on feeder layers. The measured mean stiffness of XHEF, XHFF, and MEF were 1.81 kpa, 3.61 kpa, and 1.88 kpa, respectively. While not as soft as mouse ES cells (0.6 kpa), these feeder layers fall in the softer range of biological tissues (132). They are stiffer than brain tissue (~0.1-1 kpa), but are softer than skeletal muscle (~8-17 kpa)

91 80 and precalcified bone (~25-40 kpa) ( ). More importantly, this shows that feeder layers used to support human ES cells are many times softer than rigid plastic dishes (> 3 GPa). In our studies, the softest theoretical stiffness we were able to achieve without the PDMS becoming too adhesive to handle was 3 kpa. While our statistical models generally agree that 3 kpa substrates were better than 15 kpa for pluripotency and self-renewal there may be room for improvement if the PDMS substrates were softer. Currently, 3 kpa substrates mimic the stiffness of XHFFs more so than XHEFs or MEFs. Kibschull et al. demonstrated that XHEFs can maintain CA1 and CA2 human ES cells in a defined xeno-free medium with single cell enzymatic passaging (22). With the same conditions XHFFs poorly maintained human ES cells as they rapidly differentiated and had low proliferation rates. While this difference in performance between the feeder layers may be due to differences in biochemical signals presented by the feeder layers, biomechanical differences may also explain, in part, why XHEFs and MEFs, but not XHFFs were able to robustly maintain human ES cells using single cell passaging. With respect to ECM protein components of the optimal substrate conditions, there was not a large consistent difference in culture performance when medium and substrate stiffness was held constant. A slight self-renewal advantage may exist for the fibronectin + vitronectin substrate (condition 1) as cell numbers were higher on 3 kpa E8, plastic E8, and plastic mtesr1 for each passage than other conditions within the same group. Culturing for longer periods of time, using other human ES cell lines, or using lower concentrations of growth factors would help determine the more robust ECM protein components and/or combinations. Not surprisingly, the suboptimal substrate condition of coating collagen IV on 15 kpa substrates, which was determined by the screening models, performed poorly. Human ES cells did not attach in either mtesr1 or E8, and the culture was lost in the first passage. Furthermore, no cells attached to non-coated 15 kpa substrates. Thus, the suboptimal conditions failed to promote human ES cell survival, helping to validate the screens for discriminating positive and negative hits. Interestingly, when the collagen IV was switched on to plastic, the cells managed to survive into the third passage before detaching and dying. Currently, these culture systems in E8 are near xeno-free as murine laminin and bovine fibronectin was used. In the future, human ECM proteins should be substituted for a completely xeno-free culture. More importantly, we have shown several substrate conditions defined by statistical models to be able to maintain CA1 cells over 10 passages in E8. Some encouraging

92 81 evidence suggests that these defined and near xeno-free culture systems are more robust on soft feeder-like substrates than on conventional plastic dishes. Generating more hit conditions and further characterization of existing substrate conditions will further improve and help define an effective long-term culture system for human ES cells.

93 82 Chapter 7 Conclusions and Future Work 7.1 Conclusions The potential of human ES cells to differentiate into a wide variety of human cell types enables them to be a promising tool for drug discovery, understanding early development, and regenerative medicine. Unfortunately, the current standard human ES cell culture system whether with MEFs or Matrigel/Geltrex contains undefined animal products that are not suitable for human medicine. The translation of human ES cells and more recently, ipscs from a research tool to a therapeutic reality will require the creation of chemically defined, xeno-free culture systems. In this thesis, new biomaterials that strive to mimic the ECM proteins and substrate stiffness of feeder layers supporting human ES cells were generated using statistical models derived from screening experiments. To complete this, three main objectives for the study was set out: 1. Measure the stiffness of MEF, XHEF, and HFF feeder layers using AFM. 2. Screen for optimal combinations of ECM proteins and stiffness on PDMS substrates to promote short term self-renewal and pluripotency of human ES cells. 3. Validate optimal combinations of ECM proteins and stiffness determined from the screening experiments for long-term self-renewal and pluripotency of human ES cells. The primary outcomes from this thesis are: 1. The mean stiffness of several feeder cell types used to maintain human ES cells was substantially softer than plastic substrates and was among the softer biological cell types. 2. Developed screening methods using factorial designs to determine optimal substrate conditions from four different ECM proteins and substrate stiffness for short term human ES cell pluripotency and self-renewal. 3. Validated the screening methods by testing both optimal and suboptimal substrate conditions as defined by the screening models in long-term pluripotency and self-renewal of human ES cells. Using the optimal substrate conditions, defined and near xeno-free culture systems in E8 medium were capable of maintaining CA1 human ES cells for over 10 passages. Evidence suggested that there was a self-renewal advantage for substrate conditions on 3 kpa substrates over plastic substrates in E8 medium.

94 83 Currently there are no chemically defined xeno-free culture systems robust enough to maintain long-term human ES cell cultures across different cell lines and in different laboratories. As such, the methodology for determining new potential culture systems, using both biochemical and biomechanical factors, was developed. The primary outcomes of this proof of concept study laid the foundation for incorporating additional factors, generating more hits, and validating new culture systems with multiple cell lines as well as with ipscs. In the end, the ultimate goal is that these new culture systems will help realize the potential of pluripotent stem cells in regenerative medicine. 7.2 Future Work Characterization of PDMS substrates Characterization of the PDMS substrates including empirically measuring the stiffness and quantifying the amount of protein that is adsorbed onto the substrate will help validate the culture system. Previously we had tried to measure the stiffness with AFM in order to be consistent with AFM measurements of feeder cells. However, the PDMS substrates < 9 kpa were too adhesive and sticky for indentation to obtain an accurate measurement. If the material could be altered in a way to reduce its adhesiveness, softer compositions can be measured and used for culturing. It may also be of interest to measure the stiffness of human ES cells with AFM, to see if their intrinsic stiffness is similar to feeder layers. Additionally, quantifying the amount of protein that is adsorbed onto the substrate compared to on plastic may help explain differences observed in culture performance. The relative amount of protein that adsorbs onto the substrate may vary based on stiffness and the amount of protein incubated. Preliminary data indicated that the greater the concentration of fluorescently-labeled fibronectin applied to the PDMS substrates, the greater the amount adsorbed (Figure A5). However, as PDMS substrate stiffness increased from 3 kpa to 15 kpa, less adsorbed fibronectin was detected, suggesting that fibronectin adsorption might not be independent of substrate stiffness. In contrast, the amount of adsorbed laminin generally increased with the concentration applied (with weak sensitivity), but appeared to be independent of stiffness. However, these experiments were preliminary and without replicates, and therefore further testing will be required to characterize the PDMS substrates.

95 Further characterization of new and current hit conditions The optimal substrate conditions should be cultured with additional human ES cell lines and ipscs to test their ability to robustly support other cell lines for an extended period of time. Likewise, current hit conditions should be cultured for as long as possible, while measuring additional pluripotency markers such as Sox2, SSEA3, SSEA4, TRA-1-60, and TRA-1-81 with flow cytometry for improved evaluation of the whole cell culture population. After maintaining the cells for an extended period of time, their differentiation potential can be evaluated through embryoid body formation and teratomas assays. Furthermore, karyotyping the cells is necessary to determine if any chromosomal aberrations have appeared after extended culturing. New hit conditions can be generated by repeating the screens with new cell lines or with new factors such as different ECM proteins or with human ECM proteins. Because the screens only measure the culture performance in the first passage, the amount of growth factors can be reduced in basal E8 media to create greater differences and potentially elucidate more promising optimal conditions. Reduction in ECM protein density in the screening experiments would also achieve a similar effect. Lastly, if new or current hit conditions are capable of maintaining a wide variety of cell lines over an indefinite time period, growth factor reduced media should similarly be tested to reduce media complexity and high dependence on growth factors.

96 85 References 1. Thomson, J. A., Itskovitz-Eldor, J., Shapiro, S. S., Waknitz, M. A., Swiergiel, J. J., Marshall, V. S., and Jones, J. M. (1998) Embryonic stem cell lines derived from human blastocysts, Science 282, Itskovitz-Eldor, J., Schuldiner, M., Karsenti, D., Eden, A., Yanuka, O., Amit, M., Soreq, H., and Benvenisty, N. (2000) Differentiation of human embryonic stem cells into embryoid bodies compromising the three embryonic germ layers, Mol Med 6, Liew, C. G., Moore, H., Ruban, L., Shah, N., Cosgrove, K., Dunne, M., and Andrews, P. (2005) Human embryonic stem cells: possibilities for human cell transplantation, Ann Med 37, Weick, J., Liu, Y., and Zhang, S.-C. (2011) Human embryonic stem cell-derived neurons adopt and regulate the activity of an established neural network, Proceedings of the National Academy of Sciences of the United States of America 108, Oshima, K., Shin, K., Diensthuber, M., Peng, A., Ricci, A., and Heller, S. (2010) Mechanosensitive hair cell-like cells from embryonic and induced pluripotent stem cells, Cell 141, Ng, S., Narayanan, K., Gao, S., and Wan, A. (2011) Lineage restricted progenitors for the repopulation of decellularized heart, Biomaterials 32, Rios, A., Serralbo, O., Salgado, D., and Marcelle, C. (2011) Neural crest regulates myogenesis through the transient activation of NOTCH, Nature 473, Woll, P., Grzywacz, B., Tian, X., Marcus, R., Knorr, D., Verneris, M., and Kaufman, D. (2009) Human embryonic stem cells differentiate into a homogeneous population of natural killer cells with potent in vivo antitumor activity, Blood 113, Allan, L. E., Petit, G. H., and Brundin, P. (2010) Cell transplantation in Parkinson's disease: problems and perspectives, Curr Opin Neurol 23, Naegele, J. R., Maisano, X., Yang, J., Royston, S., and Ribeiro, E. (2010) Recent advancements in stem cell and gene therapies for neurological disorders and intractable epilepsy, Neuropharmacology 58, Sumi, S. (2011) Regenerative medicine for insulin deficiency: creation of pancreatic islets and bioartificial pancreas, J Hepatobiliary Pancreat Sci 18, Kaestner, K. H. (2010) From embryonic stem cells to beta-cells: new hope for diabetics?, Gastroenterology 138, Kehat, I., Khimovich, L., Caspi, O., Gepstein, A., Shofti, R., Arbel, G., Huber, I., Satin, J., Itskovitz-Eldor, J., and Gepstein, L. (2004) Electromechanical integration of

97 86 cardiomyocytes derived from human embryonic stem cells, Nat Biotechnol 22, Schwartz, S., Hubschman, J.-P., Heilwell, G., Franco-Cardenas, V., Pan, C., Ostrick, R., Mickunas, E., Gay, R., Klimanskaya, I., and Lanza, R. (2012) Embryonic stem cell trials for macular degeneration: a preliminary report, Lancet 379, Abraham, S., Eroshenko, N., and Rao, R. R. (2009) Role of bioinspired polymers in determination of pluripotent stem cell fate, Regen Med 4, Ebben, J. D., Zorniak, M., Clark, P. A., and Kuo, J. S. (2011) Introduction to induced pluripotent stem cells: advancing the potential for personalized medicine, World Neurosurg 76, Baker, M. (2011) Stem cells in culture: defining the substrate, Nat Meth 8, Thomson, J. A., Kalishman, J., Golos, T. G., Durning, M., Harris, C. P., Becker, R. A., and Hearn, J. P. (1995) Isolation of a primate embryonic stem cell line, Proc Natl Acad Sci U S A 92, Evans, M. J., and Kaufman, M. H. (1981) Establishment in culture of pluripotential cells from mouse embryos, Nature 292, Mallon, B., Park, K.-Y., Chen, K., Hamilton, R., and McKay, R. (2006) Toward xenofree culture of human embryonic stem cells, The international journal of biochemistry & cell biology 38, Martin, M. J., Muotri, A., Gage, F., and Varki, A. (2005) Human embryonic stem cells express an immunogenic nonhuman sialic acid, Nat Med 11, Kibschull, M., Mileikovsky, M., Michael, I. P., Lye, S. J., and Nagy, A. (2011) Human embryonic fibroblasts support single cell enzymatic expansion of human embryonic stem cells in xeno-free cultures, Stem cell research 6, Ellerstrom, C., Strehl, R., Moya, K., Andersson, K., Bergh, C., Lundin, K., Hyllner, J., and Semb, H. (2006) Derivation of a xeno-free human embryonic stem cell line, Stem Cells 24, Meng, G., Liu, S., Krawetz, R., Chan, M., Chernos, J., and Rancourt, D. E. (2008) A novel method for generating xeno-free human feeder cells for human embryonic stem cell culture, Stem Cells Dev 17, Higuchi, A., Ling, Q. D., Ko, Y. A., Chang, Y., and Umezawa, A. (2011) Biomaterials for the feeder-free culture of human embryonic stem cells and induced pluripotent stem cells, Chem Rev 111, Rodin, S., Domogatskaya, A., Strom, S., Hansson, E. M., Chien, K. R., Inzunza, J., Hovatta, O., and Tryggvason, K. (2010) Long-term self-renewal of human pluripotent stem cells on human recombinant laminin-511, Nat Biotechnol 28,

98 Braam, S. R., Zeinstra, L., Litjens, S., Ward-van Oostwaard, D., van den Brink, S., van Laake, L., Lebrin, F., Kats, P., Hochstenbach, R., Passier, R., Sonnenberg, A., and Mummery, C. L. (2008) Recombinant vitronectin is a functionally defined substrate that supports human embryonic stem cell self-renewal via alphavbeta5 integrin, Stem cells 26, Hughes, C. S., Radan, L., Betts, D., Postovit, L. M., and Lajoie, G. A. (2011) Proteomic analysis of extracellular matrices used in stem cell culture, Proteomics 11, Reubinoff, B., Pera, M., Fong, C., Trounson, A., and Bongso, A. (2000) Embryonic stem cell lines from human blastocysts: somatic differentiation in vitro, Nature biotechnology 18, Bodnar, A., Ouellette, M., Frolkis, M., Holt, S., Chiu, C., Morin, G., Harley, C., Shay, J., Lichtsteiner, S., and Wright, W. (1998) Extension of life-span by introduction of telomerase into normal human cells, Science (New York, N.Y.) 279, Evans, M., and Kaufman, M. (1981) Establishment in culture of pluripotential cells from mouse embryos, Nature 292, Lachmann, P. (2001) Stem cell research--why is it regarded as a threat? An investigation of the economic and ethical arguments made against research with human embryonic stem cells, EMBO reports 2, Williams, R., Hilton, D., Pease, S., Willson, T., Stewart, C., Gearing, D., Wagner, E., Metcalf, D., Nicola, N., and Gough, N. (1988) Myeloid leukaemia inhibitory factor maintains the developmental potential of embryonic stem cells, Nature 336, Smith, A., Heath, J., Donaldson, D., Wong, G., Moreau, J., Stahl, M., and Rogers, D. (1988) Inhibition of pluripotential embryonic stem cell differentiation by purified polypeptides, Nature 336, Yoshida, K., Chambers, I., Nichols, J., Smith, A., Saito, M., Yasukawa, K., Shoyab, M., Taga, T., and Kishimoto, T. (1994) Maintenance of the pluripotential phenotype of embryonic stem cells through direct activation of gp130 signalling pathways, Mechanisms of development 45, Niwa, H., Burdon, T., Chambers, I., and Smith, A. (1998) Self-renewal of pluripotent embryonic stem cells is mediated via activation of STAT3, Genes & Development Ernst, M., Novak, U., Nicholson, S., Layton, J., and Dunn, A. (1999) The carboxylterminal domains of gp130-related cytokine receptors are necessary for suppressing embryonic stem cell differentiation. Involvement of STAT3, The Journal of biological chemistry 274, Jonathan, S. D., Christine, P., James, A. T., and Peter, W. A. (2002) Surface antigens of human embryonic stem cells: changes upon differentiation in culture*, Journal of Anatomy 200.

99 Boyer, L., Lee, T., Cole, M., Johnstone, S., Levine, S., Zucker, J., Guenther, M., Kumar, R., Murray, H., Jenner, R., Gifford, D., Melton, D., Jaenisch, R., and Young, R. (2005) Core transcriptional regulatory circuitry in human embryonic stem cells, Cell 122, Hombría, J., and Lovegrove, B. (2003) Beyond homeosis--hox function in morphogenesis and organogenesis, Differentiation; research in biological diversity 71, Schöler, H., Ruppert, S., Suzuki, N., Chowdhury, K., and Gruss, P. (1990) New type of POU domain in germ line-specific protein Oct-4, Nature 344, Nichols, J., Zevnik, B., Anastassiadis, K., Niwa, H., Klewe-Nebenius, D., Chambers, I., Schöler, H., and Smith, A. (1998) Formation of pluripotent stem cells in the mammalian embryo depends on the POU transcription factor Oct4, Cell 95, Palmieri, S., Peter, W., Hess, H., and Schöler, H. (1994) Oct-4 transcription factor is differentially expressed in the mouse embryo during establishment of the first two extraembryonic cell lineages involved in implantation, Developmental biology 166, Niwa, H., Miyazaki, J., and Smith, A. (2000) Quantitative expression of Oct-3/4 defines differentiation, dedifferentiation or self-renewal of ES cells, Nature genetics 24, Hay, D., Sutherland, L., Clark, J., and Burdon, T. (2004) Oct-4 knockdown induces similar patterns of endoderm and trophoblast differentiation markers in human and mouse embryonic stem cells, Stem cells (Dayton, Ohio) 22, Matin, M., Walsh, J., Gokhale, P., Draper, J., Bahrami, A., Morton, I., Moore, H., and Andrews, P. (2004) Specific knockdown of Oct4 and beta2-microglobulin expression by RNA interference in human embryonic stem cells and embryonic carcinoma cells, Stem cells (Dayton, Ohio) 22, Zaehres, H., Lensch, M., Daheron, L., Stewart, S., Itskovitz-Eldor, J., and Daley, G. (2005) High-efficiency RNA interference in human embryonic stem cells, Stem cells (Dayton, Ohio) 23, Henderson, J., Draper, J., Baillie, H., Fishel, S., Thomson, J., Moore, H., and Andrews, P. (2002) Preimplantation human embryos and embryonic stem cells show comparable expression of stage-specific embryonic antigens, Stem cells (Dayton, Ohio) 20, Fong, H., Hohenstein, K., and Donovan, P. (2008) Regulation of self-renewal and pluripotency by Sox2 in human embryonic stem cells, Stem cells (Dayton, Ohio) 26, Chambers, I., Colby, D., Robertson, M., Nichols, J., Lee, S., Tweedie, S., and Smith, A. (2003) Functional expression cloning of Nanog, a pluripotency sustaining factor in embryonic stem cells, Cell 113,

100 Mitsui, K., Tokuzawa, Y., Itoh, H., Segawa, K., Murakami, M., Takahashi, K., Maruyama, M., Maeda, M., and Yamanaka, S. (2003) The homeoprotein Nanog is required for maintenance of pluripotency in mouse epiblast and ES cells, Cell 113, Hyslop, L., Stojkovic, M., Armstrong, L., Walter, T., Stojkovic, P., Przyborski, S., Herbert, M., Murdoch, A., Strachan, T., and Lako, M. (2005) Downregulation of NANOG induces differentiation of human embryonic stem cells to extraembryonic lineages, Stem cells (Dayton, Ohio) 23, Rodda, D., Chew, J.-L., Lim, L.-H., Loh, Y.-H., Wang, B., Ng, H.-H., and Robson, P. (2005) Transcriptional regulation of nanog by OCT4 and SOX2, The Journal of biological chemistry 280, Chambers, I., Silva, J., Colby, D., Nichols, J., Nijmeijer, B., Robertson, M., Vrana, J., Jones, K., Grotewold, L., and Smith, A. (2007) Nanog safeguards pluripotency and mediates germline development, Nature 450, Avilion, A., Nicolis, S., Pevny, L., Perez, L., Vivian, N., and Lovell-Badge, R. (2003) Multipotent cell lineages in early mouse development depend on SOX2 function, Genes & development 17, Masui, S., Nakatake, Y., Toyooka, Y., Shimosato, D., Yagi, R., Takahashi, K., Okochi, H., Okuda, A., Matoba, R., Sharov, A., Ko, M., and Niwa, H. (2007) Pluripotency governed by Sox2 via regulation of Oct3/4 expression in mouse embryonic stem cells, Nature cell biology 9, Niwa, H., Toyooka, Y., Shimosato, D., Strumpf, D., Takahashi, K., Yagi, R., and Rossant, J. (2005) Interaction between Oct3/4 and Cdx2 determines trophectoderm differentiation, Cell 123, Yuan, H., Corbi, N., Basilico, C., and Dailey, L. (1995) Developmental-specific activity of the FGF-4 enhancer requires the synergistic action of Sox2 and Oct-3, Genes & development 9, Nakatake, Y., Fukui, N., Iwamatsu, Y., Masui, S., Takahashi, K., Yagi, R., Yagi, K., Miyazaki, J.-I., Matoba, R., Ko, M., and Niwa, H. (2006) Klf4 cooperates with Oct3/4 and Sox2 to activate the Lefty1 core promoter in embryonic stem cells, Molecular and cellular biology 26, Nishimoto, M., Fukushima, A., Okuda, A., and Muramatsu, M. (1999) The gene for the embryonic stem cell coactivator UTF1 carries a regulatory element which selectively interacts with a complex composed of Oct-3/4 and Sox-2, Molecular and cellular biology 19, Tokuzawa, Y., Kaiho, E., Maruyama, M., Takahashi, K., Mitsui, K., Maeda, M., Niwa, H., and Yamanaka, S. (2003) Fbx15 is a novel target of Oct3/4 but is dispensable for embryonic stem cell self-renewal and mouse development, Molecular and cellular biology 23,

101 Reményi, A., Lins, K., Nissen, L., Reinbold, R., Schöler, H., and Wilmanns, M. (2003) Crystal structure of a POU/HMG/DNA ternary complex suggests differential assembly of Oct4 and Sox2 on two enhancers, Genes & development 17, Chew, J.-L., Loh, Y.-H., Zhang, W., Chen, X., Tam, W.-L., Yeap, L.-S., Li, P., Ang, Y.- S., Lim, B., Robson, P., and Ng, H.-H. (2005) Reciprocal transcriptional regulation of Pou5f1 and Sox2 via the Oct4/Sox2 complex in embryonic stem cells, Molecular and cellular biology 25, James, D., Levine, A., Besser, D., and Hemmati-Brivanlou, A. (2005) TGFbeta/activin/nodal signaling is necessary for the maintenance of pluripotency in human embryonic stem cells, Development (Cambridge, England) 132, Sato, N., Meijer, L., Skaltsounis, L., Greengard, P., and Brivanlou, A. H. (2004) Maintenance of pluripotency in human and mouse embryonic stem cells through activation of Wnt signaling by a pharmacological GSK-3-specific inhibitor, Nat Med 10, Su, A., Wiltshire, T., Batalov, S., Lapp, H., Ching, K., Block, D., Zhang, J., Soden, R., Hayakawa, M., Kreiman, G., Cooke, M., Walker, J., and Hogenesch, J. (2004) A gene atlas of the mouse and human protein-encoding transcriptomes, Proceedings of the National Academy of Sciences of the United States of America 101, Macarthur, B., Ma'ayan, A., and Lemischka, I. (2009) Systems biology of stem cell fate and cellular reprogramming, Nature reviews. Molecular cell biology 10, Maruyama, M., Ichisaka, T., Nakagawa, M., and Yamanaka, S. (2005) Differential roles for Sox15 and Sox2 in transcriptional control in mouse embryonic stem cells, The Journal of biological chemistry 280, Wiebe, M., Nowling, T., and Rizzino, A. (2003) Identification of novel domains within Sox-2 and Sox-11 involved in autoinhibition of DNA binding and partnership specificity, The Journal of biological chemistry 278, Bowles, J., Schepers, G., and Koopman, P. (2000) Phylogeny of the SOX family of developmental transcription factors based on sequence and structural indicators, Developmental biology 227, Gu, P., Goodwin, B., Chung, A., Xu, X., Wheeler, D., Price, R., Galardi, C., Peng, L., Latour, A., Koller, B., Gossen, J., Kliewer, S., and Cooney, A. (2005) Orphan nuclear receptor LRH-1 is required to maintain Oct4 expression at the epiblast stage of embryonic development, Molecular and cellular biology 25, Qiu, C., Ma, Y., Wang, J., Peng, S., and Huang, Y. (2010) Lin28-mediated posttranscriptional regulation of Oct4 expression in human embryonic stem cells, Nucleic acids research 38, Yu, J., Vodyanik, M., Smuga-Otto, K., Antosiewicz-Bourget, J., Frane, J., Tian, S., Nie, J., Jonsdottir, G., Ruotti, V., Stewart, R., Slukvin, I., and Thomson, J. (2007) Induced

102 91 pluripotent stem cell lines derived from human somatic cells, Science (New York, N.Y.) 318, Chan, K., Zhang, J., Chia, N.-Y., Chan, Y.-S., Sim, H., Tan, K., Oh, S., Ng, H.-H., and Choo, A. (2009) KLF4 and PBX1 directly regulate NANOG expression in human embryonic stem cells, Stem cells (Dayton, Ohio) 27, Takahashi, K., and Yamanaka, S. (2006) Induction of pluripotent stem cells from mouse embryonic and adult fibroblast cultures by defined factors, Cell 126, Amit, M., Carpenter, M. K., Inokuma, M. S., Chiu, C. P., Harris, C. P., Waknitz, M. A., Itskovitz-Eldor, J., and Thomson, J. A. (2000) Clonally derived human embryonic stem cell lines maintain pluripotency and proliferative potential for prolonged periods of culture, Developmental biology 227, Inzunza, J., Gertow, K., Stromberg, M. A., Matilainen, E., Blennow, E., Skottman, H., Wolbank, S., Ahrlund-Richter, L., and Hovatta, O. (2005) Derivation of human embryonic stem cell lines in serum replacement medium using postnatal human fibroblasts as feeder cells, Stem cells 23, Crook, J. M., Peura, T. T., Kravets, L., Bosman, A. G., Buzzard, J. J., Horne, R., Hentze, H., Dunn, N. R., Zweigerdt, R., Chua, F., Upshall, A., and Colman, A. (2007) The Generation of Six Clinical-Grade Human Embryonic Stem Cell Lines, Cell Stem Cell 1, Richards, M., Fong, C.-Y., Chan, W.-K., Wong, P.-C., and Bongso, A. (2002) Human feeders support prolonged undifferentiated growth of human inner cell masses and embryonic stem cells, Nature biotechnology 20, Amit, M., Margulets, V., Segev, H., Shariki, K., Laevsky, I., Coleman, R., and Itskovitz- Eldor, J. (2003) Human feeder layers for human embryonic stem cells, Biology of Reproduction 68, Kleinman, H., McGarvey, M., Liotta, L., Robey, P., Tryggvason, K., and Martin, G. (1982) Isolation and characterization of type IV procollagen, laminin, and heparan sulfate proteoglycan from the EHS sarcoma, Biochemistry 21, Vukicevic, S., Kleinman, H., Luyten, F., Roberts, A., Roche, N., and Reddi, A. (1992) Identification of multiple active growth factors in basement membrane Matrigel suggests caution in interpretation of cellular activity related to extracellular matrix components, Experimental cell research 202, Xu, C., Inokuma, M., Denham, J., Golds, K., Kundu, P., Gold, J., and Carpenter, M. (2001) Feeder-free growth of undifferentiated human embryonic stem cells, Nature biotechnology 19, Wang, L., Li, L., Menendez, P., Cerdan, C., and Bhatia, M. (2005) Human embryonic stem cells maintained in the absence of mouse embryonic fibroblasts or conditioned media are capable of hematopoietic development, Blood 105,

103 Akopian, V., Andrews, P. W., Beil, S., Benvenisty, N., Brehm, J., Christie, M., Ford, A., Fox, V., Gokhale, P. J., Healy, L., Holm, F., Hovatta, O., Knowles, B. B., Ludwig, T. E., McKay, R. D., Miyazaki, T., Nakatsuji, N., Oh, S. K., Pera, M. F., Rossant, J., Stacey, G. N., and Suemori, H. (2010) Comparison of defined culture systems for feeder cell free propagation of human embryonic stem cells, In Vitro Cell Dev Biol Anim 46, Ludwig, T. E., Bergendahl, V., Levenstein, M. E., Yu, J., Probasco, M. D., and Thomson, J. A. (2006) Feeder-independent culture of human embryonic stem cells, Nat Methods 3, Hughes, C. S., Postovit, L. M., and Lajoie, G. A. (2010) Matrigel: a complex protein mixture required for optimal growth of cell culture, Proteomics 10, Evseenko, D., Schenke-Layland, K., Dravid, G., Zhu, Y., Hao, Q. L., Scholes, J., Wang, X. C., Maclellan, W. R., and Crooks, G. M. (2009) Identification of the critical extracellular matrix proteins that promote human embryonic stem cell assembly, Stem Cells Dev 18, Hynes, R. (2002) Integrins: bidirectional, allosteric signaling machines, Cell 110, Vuoristo, S., Virtanen, I., Takkunen, M., Palgi, J., Kikkawa, Y., Rousselle, P., Sekiguchi, K., Tuuri, T., and Otonkoski, T. (2009) Laminin isoforms in human embryonic stem cells: synthesis, receptor usage and growth support, Journal of cellular and molecular medicine 13, Miyazaki, T., Futaki, S., Hasegawa, K., Kawasaki, M., Sanzen, N., Hayashi, M., Kawase, E., Sekiguchi, K., Nakatsuji, N., and Suemori, H. (2008) Recombinant human laminin isoforms can support the undifferentiated growth of human embryonic stem cells, Biochemical and biophysical research communications 375, Amit, M., Shariki, C., Margulets, V., and Itskovitz-Eldor, J. (2004) Feeder layer- and serum-free culture of human embryonic stem cells, Biol Reprod 70, Beattie, G. M., Lopez, A. D., Bucay, N., Hinton, A., Firpo, M. T., King, C. C., and Hayek, A. (2005) Activin A maintains pluripotency of human embryonic stem cells in the absence of feeder layers, Stem Cells 23, Xiao, L., Yuan, X., and Sharkis, S. (2006) Activin A maintains self-renewal and regulates fibroblast growth factor, Wnt, and bone morphogenic protein pathways in human embryonic stem cells, Stem cells (Dayton, Ohio) 24, Keiji, M., Masahiko, S., Takane, H., Toshio, F., and Kohei, M. (2002) Two major Smad pathways in TGF-beta superfamily signalling, Genes to Cells Xu, R.-H., Chen, X., Li, D., Li, R., Addicks, G., Glennon, C., Zwaka, T., and Thomson, J. (2002) BMP4 initiates human embryonic stem cell differentiation to trophoblast, Nature biotechnology 20,

104 Ying, Q., Nichols, J., Chambers, I., and Smith, A. (2003) BMP induction of Id proteins suppresses differentiation and sustains embryonic stem cell self-renewal in collaboration with STAT3, Cell 115, Vallier, L., Alexander, M., and Pedersen, R. (2005) Activin/Nodal and FGF pathways cooperate to maintain pluripotency of human embryonic stem cells, Journal of cell science 118, Inman, G., Nicolás, F., Callahan, J., Harling, J., Gaster, L., Reith, A., Laping, N., and Hill, C. (2002) SB is a potent and specific inhibitor of transforming growth factor-beta superfamily type I activin receptor-like kinase (ALK) receptors ALK4, ALK5, and ALK7, Molecular pharmacology 62, Mohammadi, M., McMahon, G., Sun, L., Tang, C., Hirth, P., Yeh, B., Hubbard, S., and Schlessinger, J. (1997) Structures of the tyrosine kinase domain of fibroblast growth factor receptor in complex with inhibitors, Science (New York, N.Y.) 276, Xu, R.-H., Peck, R., Li, D., Feng, X., Ludwig, T., and Thomson, J. (2005) Basic FGF and suppression of BMP signaling sustain undifferentiated proliferation of human ES cells, Nature methods 2, Vallier, L., Mendjan, S., Brown, S., Chng, Z., Teo, A., Smithers, L. E., Trotter, M. W., Cho, C. H., Martinez, A., Rugg-Gunn, P., Brons, G., and Pedersen, R. A. (2009) Activin/Nodal signalling maintains pluripotency by controlling Nanog expression, Development 136, Xu, R.-H., Sampsell-Barron, T., Gu, F., Root, S., Peck, R., Pan, G., Yu, J., Antosiewicz- Bourget, J., Tian, S., Stewart, R., and Thomson, J. (2008) NANOG is a direct target of TGFbeta/activin-mediated SMAD signaling in human ESCs, Cell stem cell 3, Dunn, N., Vincent, S., Oxburgh, L., Robertson, E., and Bikoff, E. (2004) Combinatorial activities of Smad2 and Smad3 regulate mesoderm formation and patterning in the mouse embryo, Development (Cambridge, England) 131, Greber, B., Lehrach, H., and Adjaye, J. (2007) Fibroblast growth factor 2 modulates transforming growth factor beta signaling in mouse embryonic fibroblasts and human ESCs (hescs) to support hesc self-renewal, Stem Cells 25, Prowse, A. B., McQuade, L. R., Bryant, K. J., Marcal, H., and Gray, P. P. (2007) Identification of potential pluripotency determinants for human embryonic stem cells following proteomic analysis of human and mouse fibroblast conditioned media, J Proteome Res 6, Lim, J. W., and Bodnar, A. (2002) Proteome analysis of conditioned medium from mouse embryonic fibroblast feeder layers which support the growth of human embryonic stem cells, Proteomics 2, Moon, R., Bowerman, B., Boutros, M., and Perrimon, N. (2002) The promise and perils of Wnt signaling through beta-catenin, Science (New York, N.Y.) 296,

105 Dravid, G., Ye, Z., Hammond, H., Chen, G., Pyle, A., Donovan, P., Yu, X., and Cheng, L. (2005) Defining the role of Wnt/beta-catenin signaling in the survival, proliferation, and self-renewal of human embryonic stem cells, Stem cells (Dayton, Ohio) 23, Ludwig, T. E., Levenstein, M. E., Jones, J. M., Berggren, W. T., Mitchen, E. R., Frane, J. L., Crandall, L. J., Daigh, C. A., Conard, K. R., Piekarczyk, M. S., Llanas, R. A., and Thomson, J. A. (2006) Derivation of human embryonic stem cells in defined conditions, Nature biotechnology 24, Levenstein, M., Ludwig, T., Xu, R.-H., Llanas, R., VanDenHeuvel-Kramer, K., Manning, D., and Thomson, J. (2006) Basic fibroblast growth factor support of human embryonic stem cell self-renewal, Stem cells (Dayton, Ohio) 24, Li, Y., Powell, S., Brunette, E., Lebkowski, J., and Mandalam, R. (2005) Expansion of human embryonic stem cells in defined serum-free medium devoid of animal-derived products, Biotechnology and bioengineering 91, Hakala, H., Rajala, K., Ojala, M., Panula, S., Areva, S., Kellomäki, M., Suuronen, R., and Skottman, H. (2009) Comparison of biomaterials and extracellular matrices as a culture platform for multiple, independently derived human embryonic stem cell lines, Tissue engineering. Part A 15, Ekblom, P. (1996) Receptors for laminins during epithelial morphogenesis, Current opinion in cell biology 8, Liu, Y., Song, Z., Zhao, Y., Qin, H., Cai, J., Zhang, H., Yu, T., Jiang, S., Wang, G., Ding, M., and Deng, H. (2006) A novel chemical-defined medium with bfgf and N2B27 supplements supports undifferentiated growth in human embryonic stem cells, Biochem Biophys Res Commun 346, Baxter, M. A., Camarasa, M. V., Bates, N., Small, F., Murray, P., Edgar, D., and Kimber, S. J. (2009) Analysis of the distinct functions of growth factors and tissue culture substrates necessary for the long-term self-renewal of human embryonic stem cell lines, Stem Cell Res 3, Pyle, A., Lock, L., and Donovan, P. (2006) Neurotrophins mediate human embryonic stem cell survival, Nature biotechnology Chen, G., Gulbranson, D., Hou, Z., Bolin, J., Ruotti, V., Probasco, M., Smuga-Otto, K., Howden, S., Diol, N., Propson, N., Wagner, R., Lee, G., Antosiewicz-Bourget, J., Teng, J., and Thomson, J. (2011) Chemically defined conditions for human ipsc derivation and culture, Nature methods 8, Watanabe, K., Ueno, M., Kamiya, D., Nishiyama, A., Matsumura, M., Wataya, T., Takahashi, J., Nishikawa, S., Nishikawa, S.-i., Muguruma, K., and Sasai, Y. (2007) A ROCK inhibitor permits survival of dissociated human embryonic stem cells, Nature biotechnology 25,

106 Li, D., Zhou, J., Chowdhury, F., Cheng, J., Wang, N., and Wang, F. (2011) Role of mechanical factors in fate decisions of stem cells, Regen Med 6, Discher, D. E., Janmey, P., and Wang, Y. L. (2005) Tissue cells feel and respond to the stiffness of their substrate, Science 310, Giancotti, F., and Ruoslahti, E. (1999) Integrin signaling, Science (New York, N.Y.) 285, Wang, N., Tytell, J. D., and Ingber, D. E. (2009) Mechanotransduction at a distance: mechanically coupling the extracellular matrix with the nucleus, Nat Rev Mol Cell Biol 10, Yeung, T., Georges, P. C., Flanagan, L. A., Marg, B., Ortiz, M., Funaki, M., Zahir, N., Ming, W., Weaver, V., and Janmey, P. A. (2005) Effects of substrate stiffness on cell morphology, cytoskeletal structure, and adhesion, Cell Motil Cytoskeleton 60, Wells, R. (2008) The role of matrix stiffness in regulating cell behavior, Hepatology (Baltimore, Md.) 47, Liu, H., Sun, Y., and Simmons, C. (2013) Determination of local and global elastic moduli of valve interstitial cells cultured on soft substrates, Journal of biomechanics 46, Lisa, A. F., Yo-El, J., Beatrice, M., Miriam, O., and Paul, A. J. (2002) Neurite branching on deformable substrates, NeuroReport Engler, A., Griffin, M., Sen, S., Bönnemann, C., Sweeney, H., and Discher, D. (2004) Myotubes differentiate optimally on substrates with tissue-like stiffness: pathological implications for soft or stiff microenvironments, The Journal of cell biology 166, Engler, A. J., Sen, S., Sweeney, H. L., and Discher, D. E. (2006) Matrix elasticity directs stem cell lineage specification, Cell 126, Hansen, L., Wilhelm, J., and Fassett, J. (2006) Regulation of hepatocyte cell cycle progression and differentiation by type I collagen structure, Current topics in developmental biology 72, Gilbert, P., Havenstrite, K., Magnusson, K., Sacco, A., Leonardi, N., Kraft, P., Nguyen, N., Thrun, S., Lutolf, M., and Blau, H. (2010) Substrate elasticity regulates skeletal muscle stem cell self-renewal in culture, Science (New York, N.Y.) 329, Chowdhury, F., Li, Y., Poh, Y. C., Yokohama-Tamaki, T., Wang, N., and Tanaka, T. S. (2010) Soft substrates promote homogeneous self-renewal of embryonic stem cells via downregulating cell-matrix tractions, PloS one 5, e Veronica, C. (1999) One-Factor-at-a-Time versus Designed Experiments, The American Statistician 53.

107 Montgomery, D. C. (2001) Design and Analysis of Experiments, Jon Wiley & Sons, Inc., New York, NY Calve, S., and Simon, H. (2010) Extracellular control of limb regeneration, IUTAM Symposium on Cellular Liu, Y., Song, Z., Zhao, Y., Qin, H., Cai, J., Zhang, H., Yu, T., Jiang, S., Wang, G., Ding, M., and Deng, H. (2006) A novel chemical-defined medium with bfgf and N2B27 supplements supports undifferentiated growth in human embryonic stem cells, Biochemical and biophysical research communications 346, Keung, A., Asuri, P., Kumar, S., and Schaffer, D. (2012) Soft microenvironments promote the early neurogenic differentiation but not self-renewal of human pluripotent stem cells, Integrative biology : quantitative biosciences from nano to macro 4,

108 97 Appendix Table A1: The engineering and coded values of each factor used in the factorial design Laminin µg/cm µg/cm 2 3 µg/cm 2 Fibronectin 0 µg/cm µg/cm 2 3 µg/cm 2 Vitronectin 0 µg/cm µg/cm µg/cm 2 Collagen IV 0 µg/cm µg/cm 2 3 µg/cm 2 Stiffness 3 kpa 9 kpa 15 kpa Table A2: The 19 experimental runs of a fractional factorial design with three center points. The runs were ordered by stiffness for clarity, but the runs were performed randomly. Run Laminin Fibronectin Vitronectin Collagen IV Stiffness

109 98 Figure A1: JMP report of the screening experiment performed in mtesr1 and measured by ALP staining intensity.

110 99 Figure A2: JMP report of the screening experiment performed in mtesr1 and measured by Oct4+ cell count.

111 100 Figure A3: JMP report of the screening experiment performed in E8 and measured by ALP staining intensity.

112 101 Figure A4: JMP report of the screening experiment performed in E8 and measured by Oct4+ cell count.