1 I NTEGRATIVE STUDY ON THE RELATIONSHIP BETWEEN CELL CYCLE AND CELL MIGRATION By JUN YIN A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGRE E OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 201 5
2 Â© 201 5 Jun Yin
3 To my family
4 ACKNOWLEDGMENTS I would like to give my appreciation from my deepest heart to Dr. Spyros Svoronos and Dr. Yiider Tsen g, without whom I wouldn t be able to get this far. You contribute d to the success of my graduate studies in so many ways. Throughout my education, both of you set up idols to me for being great teachers, advisors, and good friends, which inspired me to co ntinue to pursue academ ic, since I am inspired to be person s like you. Being considerable for all the students, the sparkles you brought up during all our group meetings, all the conversations we had when I was frustrated by what happened in my life, you w ere always supportive and so nice. Y ou helped me to establish my enthusiasm for research, give my enough confidence to support me to go this far, which I never thought I could have. The trainings I ve received from you, the ways of the thinking I ve learnt from you, made me to be a better me. I am extremely lucky to have you as advisors in the most important part of my life. I would like to thank my committee members, Dr. Ben Koopman, Dr. Samu e l Wong, and Dr. Tanmay Lele, for their help throughout my resea rch. Dr. Koopman, I am so glad I met you at the first place when I got to University of Florida. Your kindness, your patience, and your humor made research no longer boring, instead, a joy. I really appreciate the way you le d me to think, to write, and I h ve never had more relaxed group meetings than the ones on Tuesdays. Dr. Wong, I made a wrong choice of not getting you as a committee member as early as I can, otherwise I could have learnt more from your smartness and calmness. Without your help, I would be still struggling about the statistics that I will never know whether it is right or wrong. You made statistics look more interesting, as in Ph.D. comics. Dr. Lele, thank you for your
5 generous on letting us to use the resource of your lab, and thanks for teaching me the presentation skills. Throughout my research, I had Shen Hsiu Hung give me initial trainings. He taught me the way to manage the lab, and also how to be a better self organization person. Thank you, Stephen Hugo Arce, all the conversation with you, all the discussion with you, all the help I got from you, made me grow faster, and made me to confirm what I want more. I have also worked with a bunch of m aster s s tudents and u ndergrad uate students, who turned out to be my good fr iends for life. For m aster s s tudents, I th ank Bo Wang, Peipei Tan, Peiyu W ang, Leyi Jing, Lin Bao, Qian Feng, Leyi Jing, for your trust and your great work. For Undergrad students, I thank Aaron Mody, Harris Krause, Melanie, for your fantastic help to my work, and for all the great conversations happened in between us. Finally, the most important people in my life my parents have given me their endless and constant love and care , they have been supportive with every decision I ve made in my life, and th ey are always the warmest bay for me. I also give my deep appreciation to my boyfriend, Dan Liu, for your love and support, for your company and help . I owe them my utmost love and grate.
6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ .......... 9 LIST OF ABBREVIATIONS ................................ ................................ ........................... 11 ABSTRACT ................................ ................................ ................................ ................... 12 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 14 1.1 Classification of Cell Migration ................................ ................................ .......... 14 1.2 Cell Cycle Regulation and Its Mechanism ................................ ......................... 17 1.3 Cell Cycle and Cell Migration ................................ ................................ ............ 18 2 EFFECT OF C HINESE MEDICINE ON THE MOBILITY AND CELL CYCLE OF GLIOBLASTOMA CELLS ................................ ................................ ....................... 21 2.1 Introduction ................................ ................................ ................................ ....... 21 2.2 Materials and Methods ................................ ................................ ...................... 23 2.3 Results ................................ ................................ ................................ .............. 26 2.4 Conclusion and Discussion ................................ ................................ ............... 29 3 CELL MIGRATION IS A SUBSET ACTIVITY OF THE CELL CYCLE REGULATION IN MODEL CELLS ................................ ................................ .......... 42 3.1 Method of Analyzing Cell Migration and the Effect of Displacement Time Interval ................................ ................................ ................................ ................. 42 3.1.1 Introduction ................................ ................................ .............................. 42 3.1.2 Description of Model ................................ ................................ ................ 45 3.1.3 Materials and methods ................................ ................................ ............ 51 3.1.4 Results ................................ ................................ ................................ .... 53 3.1.5 Discussion ................................ ................................ ............................... 55 3.2 The Relationship between the Cell Cycle and Cell Migration during Interphase of NIH3T3 Fibroblasts ................................ ................................ ........ 57 3.2.1 Introduction ................................ ................................ .............................. 57 3.2.2 Materials and methods ................................ ................................ ............ 59 3.2.3. Results and Discus sion ................................ ................................ .......... 66 4 CONCLUSION AND FUTURE WORK ................................ ................................ .... 87 LIST OF REFERENCES ................................ ................................ ............................... 92
7 BIOGRAPHICAL SKETCH ................................ ................................ ............................ 98
8 LIST OF TABLES Table page 3 1 Statistics (est imates and standard deviations) of the fraction of time ................. 77
9 LIST OF FIGURES Figure page 2 1 Chemical Structures of Z Ligustilide (LIG) a nd 3 n Butylphthalide (NBP) .......... 33 2 2 The addition of LIG to T98G cell cultures reduced migration capacity.. .............. 33 2 3 Neither cell prol iferation nor apoptosis explain the attenuation of wound like gap closure in the presence of LIG. ................................ ................................ .... 34 2 4 LIG treatment reduced T98G single cell migration capacity. .............................. 35 2 5 ................................ ................................ ................................ ................... 36 2 6 dependent manne r ......... 37 2 7 LIG at tested concentrations does not affect T98G apoptosis ............................ 38 2 8 LIG treatment reduced the expression levels of the Rho GTPases in T98G cells ................................ ................................ ................................ .................... 39 2 9 The nuclear intensity histogram of randomly sampled asynchronous T98G cells with LIG treatment at different concentrations ................................ ............ 40 2 10 The migratory capacity of NIH 3T3 fibroblasts cannot be affected by LIG treatment ................................ ................................ ................................ ............ 41 3 1 Migratory behavior can be classified into two different categories ...................... 78 3 2 The displacements of NIH 3T3 fibroblasts exhibit a cell cycle dependent property ................................ ................................ ................................ .............. 79 3 3 Idenfication of different cell cycle phas es ................................ .......................... 81 3 4 Cell migration modes are different in each cell cycle phase ............................... 82 3 5 The organizations of actin c ytoskeleton and focal ad hesions are associated with the cell cycle phases in NIH 3T3 fibroblasts ................................ ................ 8 3 3 6 The activity and mRNA profiles of Rho GTPases vary in different cell cycle phases ................................ ................................ ................................ ................ 84 3 7 The expressions and localizations of both p27 Kip1 and p21 Cip1 changes in different cell cycle phases ................................ ................................ .................. 85
10 3 8 Schematic diagram of the molecular interaction s mechanisms that lead the dynamics of focal adhesion and stress fiber formations and their attribution to cell migration during cell cycle progression. ................................ ....................... 86 4 1 Microtubule s distribution is cell cycle associated ................................ ................ 91
11 LIST OF ABBREVIATIONS CCD Cell c entroid d isplacement CCD 3 min C ell centroid displacements at 3 min intervals CN discrepancy C ellular nuclear discrepancy CNS Central n ervous s ystem ECM Extrace llular matrix EMT Endothelial to Mesenchymal Transition GBM Glioblastoma Multiforme IPA Isopropanol LIG Z Ligustilide LIMK LIM kinase MSD Mean s quare d isplacement NCD N uclear centroid displacement NBP n butylidenephthalide RAS Radi x Angelica sinensis ROCK Rho associated kinase
12 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy INTEGRATIVE STUDY ON THE RELATIONSHIP BETWEEN CELL CYCLE AND CELL MIGRATION By Jun Yin May 2015 Chair: Yiider Tseng Co Chair: Spyros A . Svoronos Major: Chemical Engineering Proliferation and migration are two cellular activities occur ring together in physiological ev ents such as organogenesis in the development stage and wound healing. They both are present in pathological events such as cancer metastasis and embryonic development . Targeting on the regulation of cell cycle is the mostly considered choice for drug disc overy, and cell migration is normally used as an efficient strategy for pre clinical study. Firstly, our study revealed the effect of Z ligustilide (LIG), an essential oil extract from Radix Angelica sinensis, which has broad pharmaceutical applications in treating cardio vascular diseases and ischemic brain injur ies on Glioblastoma M ultiforme (GBM) . We found that LIG can block T98G cells at G 1 /S phase from progression at a certain dosage, and it can also migration in a dose dependent manner. Furthermore, the attenuation of cellular mobility can be linked to the activity of the Rho GTPases (RhoA, Rac1 and Cdc42), the three critical molecular switches governing cytoskeleton remodeling; thus, regulating cell migration. LIG sign ificantly reduces the expression of RhoA and affects in a milder manner the expression of Cdc42 and Rac 1.
13 However, the spontaneous reactions of cell cycle regulation and cell motility to the same external stimuli enable us to think wh ether these two cellu lar activities cell cycle and cell migration have a strong relationship and, if so, how they regulate each other to adjust the cellular outcome has not been investigated. In our latter study, we deciph ered the cell migration mode using a model cell line NIH 3T3 fibroblasts in each stage of the cell cycle phase (the mitotic cells are not included) by the displacement differences between the cell centroids and the coupled nuclear centroids. The results suggested that fibroblasts actively migrate in G 1 pha se, stay still in S phase and possess a limited migratory activity in G 2 phase. The migratory mode changes throughout the cell cycle are highly mediated by the functions of p21 Cip1 and p27 Kip1 and focal adhesion kinase through the bridging of Rho GTPases . Together, cell migration might be a subset activity of the cell cycle in normal migratory cells; and the temporal relationship of the interactions among the proteins governing the cell cycle progression and those governing cell migration modes can be appli ed to analyze the pathological conditions.
14 CHAPTER 1 INTRODUCTION 1.1 Classification of Cell Migration Cell Migration is the process of cell translocation adopting different motility modes, such as mesenchymal, amoeboid or collective migration. It is observed in unicellular organism, and is also involved in the development of maintenance of multicellular organisms, immune responses and pathological conditions. During the processes of embryonic development, cell migration plays a vital role to guarantee the right positioning of proper cells during organogenesis. Any defect in the series of cell translocations events would lead to drastic consequence, e.g., neurological disorders, mental and physical retardation, or congenital heart disease [ 1 ] . During cancer metastasis, individual cells would detach from the primary tumor, enter lymphatic vessels for blood stream and seed in distant organs with amoeboid or mesenc hymal type movement. Or cells can adapt to collective migration by detaching the primary tumor as cel l sheets, strands and clusters [ 2 ] . There are two types of cell migration modes incorporating with each other to achieve lots of physiology process. One is Collective migration mode. It is whe n cells remain physically and functionally connected during movement. This process involves multicellular polarity and actin cytoskeleton generated traction and protrusion force to maintain cell cell junctions during cell movement [ 3 ] . Cell cell interaction and cell extracellular matrix interaction contribution to different forms of collective migration, further serve different purposes. They either come across the basement membrane, such as cells during wound closure; or they just comprise different cell lines that move together following by the formation of a ductal tree.
15 Also, there is another model excluding the cell cell context, single cell migration model. Due to the facts that single cell s are flexible enough to position themselves in tissues or secondary growth (e.g., during morphogenesis and cancer), and single cell migration also contributes to transiently pass through the tissue (e.g. immune cells), among all the studies on cell migration modes, sing le cell migration now is considered to be the best studied mecha nism of cell movement in vitro. To explicitly look at the process of cell movement, it is not hard to notice that c ells exhibit their nature of migratory behaviors by achieving the following 3 steps of events: 1) Creation/reinforce of cell substrate adhesions by generating a lamellipodia or filopodia; 2) Alignment of nuclear location and cell leading front to achieve cell polarization; 3) Detachment of cell body from the previous anchorage site s, which is generally the trailing end of the cell [ 4 ] . The polarization of cells can be initiated by external factors, such as growth factors, chemical gradients , o r extracellular matrix components leading to (de)activation of signal signaling complexes. Then the polarization is maintain and regulated by integrins, which contributes to cell matrix adhesion, following by actin re organization. Thus, the extended membr ane protrusion is formed, leading to cell body translocation. Meanwhile, over a certain time interval, the formation of focal complexes is generated at the binding sites of lamellipodia and filopodia to Extracellular matrix ( ECM ) , and then gets to maturati on. The rear trailing end of the cells will then detach from the ECM toward the cell body, pushed by the traction forces. This step also leads to cell body translocation [ 5 ] . The cytoskeleton re organization, cell substrate adhesion, and cell body contraction, are usually initiated in response to the change of signals on from extracellular matrix, or neighboring cells. Many of intracellular signal molecules have
16 been indicated to contribute to cell migration, such as small GTPases, Ca 2+ regulated proteins, mitogen activated protein kinase (MAPK) cascades, tyrosine kinases, and so on. During all the underlying signal pathways co ntributing to cel l migration, Rho family GTPases, consist of RhoA, Rac1 and Cdc42, have been studied since 1995 [ 6 8 ] , and are considered to be the most powerful swi tch involved in cell substrate adhesion, cell cell adhesion, protein secretion, and transcription. Previous studies demonstrated complexity of the crosstalk between these three molecules [ 9 11 ] . There is also increasing evidence that Rho Proteins affect the organization of the microtubule, and intermediate filaments networks, which are vital t o cell migration and also cell mitosis. Regardless of the movement modes cel ls exhibit, the sensitivity of cell migration quantification allows researcher s to be able to use it as an effective tool to identify the dynamic response of cells . It can be used for sense the change in microenvironment o r gene interference, therefore to aim to characterize the cytotoxicity of an investigation al compound in a new developed drug or gene therapy. In the past 20 years, there are tremendous investigations on understanding the molecular mechanism and pathophysiology of human disease for the dis covery of new drug. However, only very few of these fundamental discovery ended up with clinical success due to the fact that a lot of new drug selected for clinical trial are mainly previously tested in vitro on the basis of their cytotoxicity profile [ 12 , 13 ] . It is noticeable that cancer kills patients essentially due to the nature of cell migration itself. Theref ore, tools being able to characterize the dynamic response, such as anti migratory effects should draw attention as an alternative means strategy dealing with identification of new anti cancer agents or gene therapy [ 14 , 15 ] .
17 1.2 Cell Cycle Regulation and Its Mechanism Cell cycle is the series of events that occur in a cell, involving chromosome duplication and protein assembling, leading to cell division or duplication. The product of a completed cell cycle is two daughter cells after experiencing the process of interphase, mitotic phase and cytokinesis. Interphase weights the most in the scale of time, and it is spaced by intervals of growth (G 1 phase) and reorganization (G 2 phase) . In between G1 and G 2 phase, it is when the dupli cation of DNA happens (S phase). Due to its responsibility for checkpoint control, cell cycle regulation contributes to oncogenesi s and apoptosis. Meanwhile, c ell cycle is usually mentioned together with proliferation, which leads to its responsibility for cancer development. Cell cycle machinery is the key for cell proliferation control, and the occurrence of improper cell prolifera tion will eventually lead to cancer. Two sets of mechanisms of cell cycle control are recognized: 1) (De) activation of necessary protein that relay cell from one phase to another. For example, to fine tune the activity of cyclin dependent kinases ( CDK ) c yclin complexes by regulating (de)phosphorylation guarantees the smooth transitions between different cell cycle phases. 2) Check point control which senses flaws in the the pivotal events in cell cycle progression, such as DNA replication and chromosome s egregation [ 16 ] . Signals transduced by the activation of checkpoint lead to the delay of cell cycle progression, until the cells rescue themselves from the danger of becoming mutants . As mention ed above, the key parts t o the cell cycle machinery are the CDKs and cyclins, making them and CDKs related molecules under the spotlight for the selection of target of therapeutics. So far, at least 9 CDKs (1 9) and cyclin A T are known for their existence in cell cycle regulation [ 17 , 18 ] . Among these kinases and cyclins, Cyclins D1 D3/cdk2, 4,
18 and 6 contribute to G 1 phase progression; cyclin E/cdk2 controls the checkpoint into S phase to guarantee the correct timing, following by cyclin A/cdk2 controls S phase progression ; cyclin A / cdk1 (also known as cdc2) the progression of G 2 phase; and cdk1/ cyclin B are for mitosis. Inhibitory proteins, such as CDK inhibitors p16 Ink4a , p15 Ink4b , and p19 Ink4 for CDK2, 4, and 6, and kinase inhib itors p21 Waf1 , p27 Kip1 , and p57 Kip2 for blocking cyclin E/CDK 2 and cyclin A/CDK2 complexes are also considered to control cellular f unctions including cell cycl e entry and exit closely. Over or under expression will cause the disturbance of checkpoint integrity. For example, loss of p16 results to the tumor development in lung, colorectal, and breast [ 19 ] . Over expression of cyclin D1 is known to accelerate breast cancer development [ 20 , 21 ] . All these facts provide insight in developing drugs targeting cell cycle. Some of them have already entered clinical trials. 1.3 Cell Cycle and Cel l Migration The coordination of various cellular activities lays the foundation of cell physiology and enables a cell to perform its essential role in a living organism. Among different cell activities, proliferation and migration often occur as two relate d cellular events. During morphogenesis in vertebrae development, cells continuously proliferate to increase their population before they migrate to establish a desired tissue pattern [ 2 2 ] . Similarly, endothelial and smooth muscle cells consequentially proliferate and migrate to repair wound appearing in coronary, peripheral and cerebral arteries [ 23 ] . Further, cancer cells grow and sequentially invade/migrate to metastasize to remote sites regardless the contents of their mutations and dysregulations [ 24 ] . Even though migration and proliferation do not necessarily occur simultaneously, it is of interest to investigate whether a common relationship exists between them for them to work harmonically to achieve a physiological or pathological event.
19 Presently , it is generally accepted that the migratory state of a cell is affected by the phase of the cell cycle. Lymphocyte migration likely happens in G 1 phase rather than S and G 2 /M phase [ 25 ] . The motilities of L929, HeLa and BT4Cn glioma cells are greater in G 1 /S phases compared to G 2 /M phases [ 7 ] . In addition, fibroblast growth factor 1 (FGF 1) can induce the movement of NBT II rat bladder carcinoma cells in G 1 phase [ 23 ] . However, it is unclear whether these are independent cases or, rather, a common molecular mechanism exists to integrate the cell cycle and cell migration. If these two cellul ar activities are indeed tightly connected, their relationship could connect many seemingly dispersed signaling events into a more cohesive model in biomedical studies. The decrease of cell migration activity after the cells pass G 1 /S transition in above examples might be epitomized by the special dynamic chores associated to individual cell cycle phases, including metabolism, signal transduction, etc. [ 23 , 26 ] . From physical aspects, the volume of a cell generally elevates more than two times through the interphase [ 27 , 28 ] . Based on the studies in HL 60 cells, the cytoplasmic viscosity also upsurges approximately 1.5 times from G 1 to S phase [ 25 ] . In addition, the surface area of the cell is significantly spanned over the progression of the cell cycle [ 29 ] . Each of these physical properties can be used as part of the reasons to interpret why the progression of the cell cycle affects cell migration capacity regardless of the cell types. Yet, from molecular aspects, how the proteins regulating cell migration accommodate with the proteins controlling the cell cycle to alter the cell migration mo de through the natural cell cycle progression is not fully established. Cell migration is facilitated by cytoskeletal remodeling, in which Rho GTPases (i.e., RhoA, Rac1 and
20 Cdc42) serve as the primary molecular switches [ 7 , 8 ] . Meanwhile, the cyclin CDK complexes regulate the progression of the cell cycle. Hence, the crosstalk between these two groups of proteins should provide insight for the relationship between these two cellular activities in cell physiology.
21 CHAPTER 2 EFFECT OF CHINESE MEDICINE ON THE MOBILITY AND CELL CYCLE OF GLIOBLASTOMA CELLS 1 2.1 Introduction R adix Angelica sinensis (RAS) has been considered a medicinal plant and applied to alleviate various disease syndromes in traditional Chinese medicine for over a thousand years. Approximately, more than 70 compounds have been currently identified in RAS, including phthalide dimers, organic acids and their derivative esters, polyacetylenes, vitamins, amino acids, and essential oils [ 30 ] . Among the essential oils of RAS, Z ligustilide (LIG) is one of the most active co mponents and has been characterized for more than 40 years. LIG can inhibit the proliferation and cell cycle progression of vascular smooth muscle cells, associated to basic fibroblast growth factor stimulation, through the reduction of reactive oxygen sp ecies and/or the suppression of the MAPK pathway [ 31 ] . LIG also inhibits vasocons triction induced by norepinephrine bitartrate and calcium chloride on rat abdomina l aorta segments [ 32 ] . Hence, LIG is considered to be an effective agent to reduce vascular resistance; thereafter, increase blood flow and enhance microcirculatio n to prevent cardiovascular diseases , including atherosclerosis and hypertension [ 33 , 34 ] . Meanwhile, LIG has an analgesic effect on rats and a concentration dependent anti inflammatory effect on lipopolysaccharide activated rat microglia without cytotoxicity [ 35 , 36 ] . LIG is also known to have a protective effect against ischem ic brain injury caused This chapter has been published by Yin J, Wang C, Mody A, Bao L, Hung S H, et al. (2013) The Effect of Z Ligustilide on the Mobility of Human Glioblastoma T98G Cells. PLoS ONE 8(6): e66598. doi: 10.1371/journal.pone.0066598
22 by the failure of regular blood supply to local brain tissue in the central nervous system (CNS) [ 37 ] . LIG decreases the level of malondialdehyde, a product of lipid peroxidation, and increases the activity of antioxidant en zymes, fostering an anti apoptotic effect that reduces cerebral infarct volumes and improves neurobehavioral deficits [ 38 ] . The structure of LIG is similar to that of n butylidenephthalide (NBP) (Fig. 2 1), which has also been demonstrated to possess activity to reduce inflammation and hepatotoxicity as LIG does [ 30 ] . A recent study has revealed that NBP could suppress the growth of Glioblastoma Multiforme (GBM) cells both in vitro and in vivo via cell cycle arrest and a poptosis [ 39 ] . GBM is the most common and aggressive malignant primary brain tumor, represents 50 % of all gliomas and has the worst prognosis of any CNS malignancy despite the progression of existing diagnosis methods and treatments [ 39 ] . The unrevealed rapid invasion mechanism of GBM presents a great challenge to accurately predict the development of GBM and efficiently treat it. As a derivative of NBP, LIG might have similar pharmaceutical effects on GBM diseases; therefore, the pharmaceutical outcome of LIG treatment of GBM is worth investigating. The most common assessment s for drug effects on cells are endpoint evaluations, such as the induction of apoptosis and the change in cell proliferation. However, other beneficial drug effects might exist and can be tested through non conventional methods. In this study we explored the effect of LIG treatment on T98G cells not only using endpoint assessments, but also by evaluating the changes in cell migration patterns, one of the most critical cell activities of cancer metastasis. Cell migration patterns were assessed at both the cellular and the molecular level. The three Rho GTPases (RhoA, Rac1 and Cdc42) are the main molecular switches that govern
23 cytoskeletal remodeling to regulate cell migration [ 8 ] , and these Rho GTPases are usually linked to cancer by changes in their expression profiles rather than by mutations [ 8 ] . Hence, the expression levels of these three proteins were evaluated via Western blotting. 2.2 Materials and Methods Cell culture and preparation of LIG : T98G cells were obtained from ATCC (American Type Culture Collection, Manassas, VA) and cultured in EMEM (Mediatech, Manassas, VA), supplemented with 10 % fetal bovine serum (Hyclone Laboratories, Logan, UT). Cell cultures were maintained in a 5 % CO 2 incubator with routine passing every 2 ~ 3 days. Cells were transferred onto glass bottom dishes (In Vitro Scientific, Sunnyvale, CA), p re treated with 0.01 % poly L lysine (Sigma Aldrich, St Louis, MO) and 20 ÂµL / mL fibronectin (BD Biosciences, Bedford, MA), for image acquisition. LIG was isolated by silica gel column chromatography from the Angelica sinensis essential oil, which was extr acted using supercritical CO 2 fluid. LIG was then identified by 1 H and 13 C nuclear magnetic resonance spectrometry and electron impact ionization mass spectrometry as 99.8 % pure (density = 0.979 Â± 0.005 g/ mL ). Because it is insoluble in water, isopropanol (IPA) was used as an effective solvent. LIG was first diluted 1,000 , 2,500 , 5,000 and 10,000 fold in IPA, and then applied to T98G cell cultures at a further 200 fold dilution to reach final LIG concentrations of 25, 10, 5 and 2.5 ÂµM, respectively. Wo und like gap closure assay : Wound like gap closure (or wound healing) assays were performed on 6 well pre treated glass bottom dishes (In Vitro Scientific). Cells were loaded at a concentration of ~ 5 10 5 cells/ mL , and grown to confluence. A
24 glass pipet wa s used to physically scratch a ~ 400 Âµm wound gap in the monolayer of cells. Samples were then washed with HBSS (Mediatech Inc.) and then incubated in fresh culture media before the following additions: 1) mock (negative control); 2) 0.5 % IPA; 3) 2.5 ÂµM L IG; 4) 5 ÂµM LIG; 5) 10 ÂµM LIG; and 6) 25 ÂµM LIG. Here, the sample with 0.5 % IPA also served as a control to monitor the effect of the solvent of LIG at the same applied concentration on cells. Cell motility was recorded using a Nikon TE200E microscope sys tem (Nikon Instruments, Melville, NY), equipped with a 10 objective lens and an EMCCD camera (Roper Scientific, Tucson, AZ). Six positions along the scratched areas under each condition were recorded at several time points during a 14 hour period for furt her analysis. Quantification of wound like gap closure process : A gap index ( GI ) was employed using the ratio of remaining area to the original wound gap area utilizing a custom algorithm coded in Matlab (Mathworks, Natick, MA) . Each 1000 pixel Ã— 1000 pix el image at the initial time, t 0 , was divided into 2500 20 pixel Ã— 20 pixel boxes and the average intensity of each box was obtained. Twenty boxes without cells were picked to determine the average background intensity, and then three standard deviations a bove the average background intensity was set as a threshold value. The GI of selected areas at time t was calculated by the following equation: Cell proliferation and apoptosis assay T98G cells at a concentration of ~1 10 5 cells/well were plated in 12 well culture dishes (Corning Inc., Corning, NY). Cells in media with either no treatment, 0.5 hemocytometer at different time points until cell concentration was doubled. Triplicate
25 tests at each treatment condition were conducted, and each of the triplicates was measured 3 tim es. Cell counts vs. time were plotted and fitted into exponential curves. The APO DIRECT KIT (BD Biosciences) was used to assess the apoptotic state cell concentrations corresponding to wound like gap closure assay and single cell assay, respectively. T98G cells treated with 8 M camptothecin (Sigma Aldrich) for 20 hours were used as a positive control. Samples were assessed using an ACCURI C6 Flow Cytometer (BD Bioscience s). Single cell assay and data analysi s : The pEGFP plasmids (BD Biosciences) were introduced into T98G cells at approximately 50 % 70 % confluence by transfection. The transfected cells were re plated on fibronectin coated 6 well glass bottom plates (In Vitro Scientific) at a concentration low e nough to avoid cell cell expressing green fluorescence protein were monitored every minute for 1 hour with a 20 objective lens under a UV light source (X Cite 120 PC fluore scent light source, EXFO, Ontario, Canada). Cell movement videos were analyzed using the MATLAB software image processing toolbox. Western blotting: Samples were prepared by seeding T98G cells at high (~5 10 5 cells/ mL ) and low (~5 10 3 cells/ mL ) concentrati ons, corresponding to wound like gap closure assay and single cell assay, respectively. Cells prepared at high probe the individual activities of the Rho GTPases in the w ound like gap closure process under different treatment conditions. Cells at low concentration were incubated
26 proteins in single cell assay conditions. Also, low densi ty cells with the same treatments were incubated for 20 hours for cross comparison purposes. Each individual culture sample was suspended in 300 ÂµL lysis buffer (Cytoskeleton, Denver, CO) with 1 % of protease inhibitors (Cytoskeleton) and homogenized on ic e. After centrifugation at 4 C, the supernatants of the cell lysates were separately collected for protein detection. Total protein concentration of a supernatant was measured using the Coomassie Plus Protein Assay (Thermo Scientific, Waltham, MA) and ~ 8 Âµg of sample was loaded on a 12 % polyacrylamide gel for electrophoresis, followed by transference onto a PVDF membrane (BIO RAD, Hercules, CA). Individual Rho GTPases were detected with the specific monoclonal antibody against Rac1 (Millipore, Billerica , MA), RhoA (Santa Cruz Biotech Inc., Dallas, TX) or Cdc42 (BD biosciences). Here, GADPH (Santa Cruz Biotech Inc.) activity was used as a baseline control for all the blotting. Results of western blotting were quantified using ImageJ (NIH, Bethesda, MD), t reating proteins in untreated T98G cells as reference. 2.3 Results LIG exhibits a cellular effect hindering the migration activity of T98G cells : To explore whether LIG treatment affects the migration of T98G cells, LIG was diluted to different concentrat ions in IPA and applied to T98G cell cultures immediately after the wound like gap was created. The progression of wound like gap closure was then recorded and analyzed. The results suggest that wound closure capacity decreases in the presence of LIG but n ot for IPA treatment (Fig. 2 2 A). Moreover, this LIG effect was
27 dose dependent (Fig. 2 2 B). Based on these results, 5 ÂµM is the lowest dose of LIG that has significant effect on cell migration; so we chose this dosage for further study. The migration hin dering effect was not caused by reduction in cell proliferation rate or apoptosis : Changes in cell proliferation alter the cell population and may affect the wound the proliferation rate of T98G cells was investigated. A hemocytometer was employed to count cell populations wit h or without the presence of LIG over a 20 hour period. Cell populations over the monitored time periods were fit by exponential curves to obtain the growth equations and regression coefficients (Fig. 2 3 A). Population growth rates suggest that LIG at the tested concentration did not lead to any detectable difference on cell proliferation (specific growth rates: 0.0263 hr 1 for T98G, 0.0264 hr 1 for T98G + IPA and 0.0263 hr 1 for T98G + LIG). Hence, the reduction of wound like gap closure capacity of T98G cells was not due to a change in cell proliferation. Cell apoptosis could also affect the cell population during the wound like gap closure process. Hence, the ability of LIG at the working concentration to induce cell apoptosis was examined on T98G cell s of both low and high density. The percentage of cell viability (2 trials) indicated that LIG does not induce T98G cell apoptosis at both high and low cell concentration (Fig 2 3B; also see Fig. 2 8 ). Thus, the LIG effect that hinders the wound like gap c losure process was not due to induction of apoptosis. LIG treatment affects single T98G cell migration : Cell cell interactions can also affect the progression of wound like gap closure. The interplay between the signaling pathways governing cell migration and cell cell interaction has been investigated intensely, especially the topic of endothelial to mesenchymal transition
28 (EMT) related to cancer metastasis [ 40 ] . Therefore, cell cell intera ction should not be overlooked as a potential factor that influences wound like gap closure. To that end, the effect of LIG on the mobility of individual T98G cells was studied. Fluorescence images of single T98G cells expressing green fluorescence protei n (GFP) were captured every minute for one hour to track their movement after the cells cell centroid was determined and the cell centroid displacements (CCD) between ima ges were calculated and combined to form single cell trajectories (Fig. 2 4 A). The lengths of T98G cell trajectories indicate that 0.5 % IPA might also affect cell migration at the applied concentration. Further scrutiny of the single cell migration by me an square displacement (MSD) vs. time interval in log log plots revealed that the MSD of the T98G cells was not appreciably affected by 0.5 % IPA (Fig. 2 4 B, first two panels). MSD was clearly reduced to ~ 0 (Fig. 2 4 B, the third panel). The histograms of the r five hours noticeably reduced the occurrence of cell step sizes larger than 0.5 Âµm (Fig. 2 4 C, all panels and insets). LIG treatment changes the expression levels of Rho GTPase : Biophysical assessments provide direct and clear evidence for the LIG effe ct on T98G cell migration. We further investigated the molecular cause of the LIG effect by monitoring the expression levels of proteins directly associated with cell migration. Since the Rho GTPases, RhoA, Rac1 and Cdc42, are known as the full set of mole cular switches that collectively govern cytoskeletal remodeling during cell migration, Western blotting
29 against those three Rho GTPases was conducted to provide a molecular explanation for the observed LIG effect. The Western blotting results at high cell density, with or without 20 hour LIG incubation, were compared to probe the LIG effect on T98G cells in the wound like gap closure process (Fig. 2 5 A). Samples at low density (single cell assay conditions), with or without 5 hour LIG incubation, were als o studied (Fig. 2 5 B). The 5 hour incubation time was chosen because this is the time frame when single migrant cells were observed to leave the bulky cell cluster during the wound like gap closure process. To make it easier to cross compare results from these two conditions, Western blotting was also conducted at single cell concentrations with 20 hour LIG treatment (Fig. 2 8 ). The results elucidated that the expression levels of the three Rho GTPases were all modulated under LIG treatment and that RhoA w as decreased the most. These results were in agreement with the conclusion drawn from the biophysical assays: the LIG treatment effect was due to the presence of LIG and not 0.5 % IPA. Hence, the LIG effect to reduce T98G cell migration ability was confirm ed. 2.4 Conclusion and Discussion This study demonstrates that Z ligustilide (LIG), an extract from the essential oils of RAS, has a specific effect on GBM T98G cells to hinder cell migration. This conclusion is supported by consistent results from three d istinct approaches. 1) The monitoring of the wound like gap closure process with the consideration of apoptosis and changes in cell proliferation. 2) Direct single cell migration analysis, which supports the wound like gap closure results and rules out cel l cell interactions as a cause of reduced migration. 3) The measurement of the expression levels of Rho GTPases, the
30 three master regulators of cell migration. The wound like gap closure assays also suggest that this effect is dose dependent, which further strengthens our conclusion. It has been reported that LIG has a cytotoxic effect on certain cell types, such as PC12, SH SY5Y, HeLa, and C6, when the concentration of LIG is higher than 50 ÂµM [ 41 ] . We also tested the effect of LIG on arresting cell cycle of T98G at the recommended concentration, as shown in Fig. 2 9. LIG wit h the concentration of 25 ÂµM or higher could hind er the progression of cells from S to G 2 phase. It is notable that at the concentration of 25 ÂµM , LIG was able to reduce migration capacity of T98G cells significantly. As a control cell line, NIH 3T3 fibroblasts were treated with the working concentration , 5 ÂµM . No significant different were detected between the groups before and after LIG application , as shown in Fig. 2 10 . Interesting ly , this observation is also consistent with the study by Xia et. a l [ 42 ] , which indicate d that no significant effect of Ligustilide was observed on normal mouse fibroblasts. This study shows that LIG can significantly change the migration pattern of T98G cells at much l ower concentration than 5 ÂµM. Besides the identification of the new pharmaceutical effect of LIG treatment on T98G cells, this study also demonstrates that cell migration analysis can be a highly sensitive method for detecting cellular effects of potential drug candidates, which can be difficult to probe by endpo int assessment in pharmacology. This also explains the concentration that works on hindering cell migration might have non significant effect on cell cycle arre st. However, it also indicates that th e effects of the test drug on cell migration and cell cycle are spontaneous .
31 IPA is a non polar solvent that is used to dissolve LIG in the mobile phase for HPLC to separate LIG from other RAS extracted compounds [ 43 ] . LIG dissolved in IPA can be homogeneously distributed to cell media. However, our study shows that IPA alone could slightly alter the single cell migration pattern but not the wound like gap closure process involving cell cell interactions. Hence, the presence of IPA may direc tly affect cell behavior by hindering the cell migration capacity but this effect could be shielded by cell cell interactions through signaling crosstalk. There is another possibility: the evaporation of IPA during the incubation period could also make the inconsistent. Since the wound like gap closure process takes much longer (more than 14 hours) than the single cell assay (only monitored for one hour during the fifth hour iminished and down played during the time course of the wound like gap closure process. A single cell is subjected to fewer constraints when it attempts to change moving direction compared to cells in clusters. This freedom comes from not only the spatial availability but also, more predominantly, from the governance of cell polarity by cell cell and cell substrate interactions [ 40 ] . The signaling proteins involved in cell cell interactions m ight also influence cell migration. Hence, the single cell migration analysis could provide more straightforward information regarding cell migration because several factors that could affect the wound like gap closure process (i.e., cell cell interactions , cell density and the rates of cell apoptosis and proliferation) could be excluded from single cell migration analysis. When the results from the wound like gap closure process and single cell migration are compared, they suggest that the effect of LIG is significant.
32 The three Rho GTPases, RhoA, Rac1 and Cdc42, are known to be the molecular switches for the actin cytoskeleton [ 8 ] remodeling. Our results further sug gest that the LIG effects on cell activities are through fine tuning the expression levels of Rho GTPases. It has been reported that it is the expression levels of Rho GTPases, but not their mutations, that govern change of cancer cell activities, such as metastasis. Furthermore, it has been shown that LIG can inhibit ERK MAP kinase, which is an upstream regulator of RhoA and Rac1 in colon carcinoma cells [ 44 ] , and s uppress tumor necrosis factor alpha (TNF pulmonary microvascular endothelial cells [ 45 ] . Moreover, in PC12 cells, LIG can activate PI3K/Akt pathways [ 46 ] , which directly up regulate Rho GTPases. The study provides a clear perspective that explains the inhibitive effect of Z ligustilide on T98G cell migration and suggests that Z ligustilide may be an effective agent for protecting the central nervous system against GBM diseases, parti ally through the modulation of Rho GTPase expression. Furthermore, this study also shows that cell migration can be used to understand the outcome of altered Rho GTPase activity as a potential approach for drug screening.
3 3 Fi gure 2 1. Chemi cal Structures of Z Ligustilide (LIG) and 3 n Butylphthalide (NBP) F igure 2 2 . The addition of LIG to T98G cell cultures reduced migration capacity. A) Images of the wound like gap closure process of T98G cells under mock, 0.5 % isopropanol (IPA), 5 ÂµM LIG, and 25 ÂµM LIG (top to bottom) are presented at the 0 th , 8 th and 14 th hour. The yellow lines indicate the approximate boundary between cell inhabited and cell free (central) regions of each image. Scale bar: 200 Âµm. B) The result of a quantitative ana lysis of the gap closure of T98G cells under no treatment , 0.5 % IPA, and 4 concentrations (2.5, 5, 15 and 25 ÂµM) of LIG treatment is shown. The gap index is the percentage of original wound gap that remains cell free. The data are the mean gap index resul ts of 6 experiments with error bars representing the standard deviation (also refer to Fig. 2 6 ).
34 Figure 2 3. Neither cell proliferation nor apoptosis explain the attenuation of wound like gap closure in the presence of LIG. A) Comparative growth of T98G cells under no treatment, 0.5 % IPA and 5 ÂµM LIG treatment is shown. Each population of T98G cells was counted every 3 hours for 24 hours. The average number from 9 assessments at each time point was fit to an exponential curve to obtain the growth rate u nder each treatment condition. Error bars represent the standard deviation. B) The result of apoptosis assays for both high and low density of T98G cells is shown. The bars represent the average between two measurements of viable cell percentages under no treatment, 0.5 % IPA, and 5 ÂµM LIG treatment, respectively.
35 Figure 2 4. LIG treatment reduced T98G single cell migration capacity. A) Trajectories of T98G cells indicate that the presence of LIG could drastically hinder cell migration, and the additio n of 0.5 % IPA has a minor effect. B) The mean square displacement (MSD) vs. time interval is shown as a log log plot, demonstrating that LIG treatment significantly decreases the slope of the MSD of T98G cells as compared to control conditions. The mean M SD value for each condition is plotted as a dotted black line. C) Histograms of the one minute step sizes within one hour suggest a reduction of the translocation of T98G cells under LIG treatment since the occurrence of larger step sizes is greatly reduce d.
36 F igure 2 5
37 Figure 2 6. dependent manner. The wound like gap closure process of T98G cells unde r mock, 0.5 % isopropanol (IPA), 2.5 ÂµM, 5 ÂµM, 10 ÂµM, and 25 ÂµM LIG (top to bottom) is presented at 0 th , 3 rd ,8 th , 14 th and 18 th hour until the mock T98G totally healed. Yellow lines indicate the approximate boundary between the cell free wound region of images and the region corresponding to migrating cells. Scale bar: 200 Âµm.
38 Figure 2 7 . LIG at tested concentrations does not affect T98G apoptosis. Three groups of experiments were conducted: A) T98G cells, treated with and without 8 M camptothecin , were tested through flow cytometry to achieve the selection of population gate and apoptosis gate ((P3: population gate for selecting cells to be analyzed R1: apoptosis gate for selecting apoptotic cells). B) High density T98G cells and C) Low concentrat ion T98G without any cell cell interactions were each incubated with 0.5 % IPA and 5 ÂµM LIG solution, respectively. LIG did not lead to any difference in the number of apoptotic cells. Each experiment was conducted in duplicate.
39 Figure 2 8 . LIG treat ment reduced the expression levels of the Rho GTPases in T98G cells. T98G cells were loaded at the conditions for single cells and incubated with 0.5 % IPA or 5 ÂµM LIG for 20 hours and then analyzed by Western blotting against RhoA, Rac1 and Cdc42 (a house keeping protein, GADPH, was also detected as a reference to to calibrate the relative amount of RhoA, Rac1 and Cdc42.). The results were normalized using the control condition (no treatment) as reference.
40 Figure 2 9. The nuclear intensity histogram of r andomly sampled asynchronous T98G cells with L IG treatment at different concentration s were plotted (Top to bottom: 0, 10, 25, 50 M, respectively). Orange d ash line indicates the peak of 1 copy DNA, which is approximately the mean of G 1 phase peak, and the black dash line indicates the approximate mean of G 2 phase peak.
41 Figure 2 10 . The migratory capacity of NIH 3T3 fibroblasts cannot be affected by LIG treatment. A) Histograms of minutely displacement of NIH 3T3 cells without (left) or with (right) L IG treatment were plotted. Test concentration was 5 M for the comparison to group of T98G cells. B) Hourly displacement of NIH 3T3 cells without (left) or with (right) LIG treatment were plotted. No significant difference was detected between two groups. Test concentration was 5 M.
42 C HAPTER 3 CELL MIGR ATION IS A SUBSET ACTIVITY OF THE CELL CYCLE REGULATION IN MODEL CELLS 3.1 Method of Analyzing Cell Migration and the Effect of Displacement Time I nterval 3.1.1 Introduction Cell migration is an evolutionarily conserved mechanism that underlies the develo pment and functioning of uni and multicellular organisms. It takes places in normal and pathogenic processes, including various events of embryogenesis, wou n d healing, immune response, cancer metastasis, and angiogenesis [ 1 ] . Dysregulation of cell migration would lead to severe malfunction of human physiology. Cell migration models can be classified into three different modes based on the cell migration mechanism s : collective cell migration, single amoeboid migration, and single mesenchymal migration . Collective cell migration could occur during processes such as morphogenesis, vascular sprouts during angiogenesis, or wound healing across wound matrix , in which cel ls remain physically connected with each other and the integrated cell property delivered by cell cell interactions is preserved . Amoeboid motility is characterized by cycles of expansion and contraction of the cell body , regulated by the cortically locali zed actin and myosin. Amoeboid type migration enable s the cells to squeeze through the gaps in the extra cellular matrix (ECM) fibers and adapt their bodies to the pre existing spaces. This type of push and squeeze migration makes amoeboid migration play a pivotal role in cancer metastasis due to the flexibility of cells. Different to the se two modes discussed above, cells with mesenchymal type of migration are more likely to be fibroblast like motility. Apart from fibroblasts, Keratinocytes, endothelial ce lls, and some tumor cells also use this type of motility.
43 M esenchymal motility begins with the extension of lamellipodia initiated by the extra cellular matrix (ECM) proteolysis . The interaction of cell body and ECM can further trigger downstream signaling events followed by the contraction of the rear end of the cell and displacement of the cell [ 47 ] . Cell migration type can be considered as a phenotypic ch aracterization of a cell line itself, thus cells may utilize various types of migration patterns to adapt to environmental changes and matrix stiffness due to drug treatments to continue migrating. For example, the amoeboid and mesenchymal types of migrati on modes are mutually interchangeable when suppression or enhancement of the specific molecular pathways occurs in respond to current specifics of environment during the invasiveness of cancers. To be able to better understand the cell migration modes, i nc reasing attention has been addressed on the experimental measurements and mathematical models to classify these modes , such as directed (mesenchymal) migration and random migration . The selection of these models is normally based on the long term documenta tion of cell trajectories. However, w hen explicitly viewing from the subcellular perspective, mesenchymal cell migration is a complex activity consisting of a dynamic extension of the leading edge toward the cell front (denoted as protrusion), and a retra ction of the trailing edge for the cell to move forward (denoted as detachment). Also, a nucleus, which comprises about 30 % of the total cell mass, has a direct impact on the capacity of mesenchymal cell migration [ 48 ] . When a single cell undergoes protrusion, the nucleus usually stays stationary. In contrast, significant cell membrane detachment is always accomplished with a similar degree of nuclear movement in the same rear retraction event for an
44 active mesenchymal migratory cell. Once the time intervals of the cell image stacks (e.g., 1 min) are shorter than the locomotion events, a trailing edge membrane detachment event often gives rise to approximately the sam e degree of nuclear displacement. In the detachment event, the whole nucleus tends to moves in a larger dimension than cells, leading to the characteristic of a greater nucleus centroid displacement (NCD) than the coupled cell centroid displacement (CCD). In contrast, a protrusion (or con traction) domina nt event can be interpreted simply with CCD NCD. Mostly, protrusion is the hallmark of the cell body translocation initiation. Cells with frequent protrusion and few detachment would highly likely possess smaller migration ability in a long term than cells with modest protrusion frequency and more detachment. The frequency of protrusion (or con traction) or detachment events can be considered as a nature feature of a certain cell type therefore further used to connect with other mechanism study or to characterize the environmental change. Therefore, the difference between cell speed and the corresponding nucleus speed is of great interest to be investigated. Although population parameters, such as persistence time, mean square displacement, or wound healing recovery rate are largely documented [ 49 ] , a method that is sensitive enough to characterize the migratory behavior (protrusion and detachment) of individual cell is not yet systematically studied. In addition, in this type of analysis, a short time interval is favored [ 50 ] . However, an additional challenge arose by using short imaging time intervals is the presence of measurement errors from the tracking system. Adopting short term analysis will also allow us to look into time sensitive environment, such as cell cycle progression, and treatment of short term efficacy drugs.
45 In this study, we develop a method to accurately estimate the fraction of time where cell centroid speed exceeds that of the corresponding nucleus since it is one of the most importance indicator of initiating migratory behavior, regardless of cells lines.We do so on a time scale that is appropriate to resolve protrusion and detachment events. The purpose of this estimation is to provide a quantitative characteristic to describe and compare the amount of sampling and migration activity migratory cells, in our study, NIH 3T3 fibroblast cells . 3.1.2 Description of Model Setup an d a discrete time approximation : In single cell movements, C ( t ) and N ( t ) denote the true positions of the cell and nuclear centroids, respectively, and s C ( t ) and s N ( t ) denote the speeds of the cell centroid (CCS) and nuclear centroid (NCS) over a unit of time , respectively. Consequently, the indicator function is set as I(A) = Then, the fraction of time where cell centroid spee d exceeds the corresponding nuclear centroid speed can be expressed as: ( 3 1) The average value of this integral over a population of cells is our primary interest. The data shall consist of measurements of cel l centroid and nucleus centroid positions at the discrete time points, t i , where the subscript i = 0, 1, m and . Without any assumptions on the dynamics of the overall cell trajectories, a simple approximation for this integral is provided by the Riemann sum
46 ( 3 2) where d(X,Y) denotes the Euclidean distance between two points X and Y. With equally spaced time intervals , equation ( 3 2) simplifies to ( 3 3) Here, is just CCD for the time interval [t i 1 , t i ]. Likewise, is NCD for the time interval [t i 1 , t i ]. Suppose is sufficiently small in the sense that a protrusion (or con traction ) or detachment event spans multiple time intervals, e.g. = 1 minute. Then, it follows that the difference in the magnitudes of NCD suggesting protrusion (or con traction ) and CCD < NCD suggesting detachment. See Figure 1 f or a schematic of this relationship. However, due to measurement error, the true values of C(t i ) and N(t i ) are not known. Instead, we observe the noise contaminated measurements C * (t i ) and N * (t i ). Assume that the errors are additive, so that we have the r epresentations and for i = 0, 1, .., m . The and represent the 2D error term in the (x,y ) coordinate plane, for cell and nucleus respectively. We also assume that the probabili ty distribution of and does not depend on the time and location of the measurements, and that the errors are all independent. With errors, the above formula becomes ( 3 4)
47 where C * (t i ), N * (t i ), C * (t i 1 ) and N * (t i 1 ) are observed quantities and each s is a random variable based on the probabilities of the measured position distributions. Let denote the average (expected value) of this quantity. Since the expected value of an indicator, I(A) , for any set A is the probability P(A) , we get ( 3 5) Thus, an approximation to integral (equation ( 1 ) ) for a single cell can be calculat ed after determining the probabilities P(CCD > NCD) when the time intervals are chosen. In the presence of measurement error, it is obvious that the approximation depends on the choice of . For this study, we chose = 1 minute. We further elaborate o n this issue in the Discussion section. We also adopt a total observation period of 1 hour per cell. Constructing the probability distribution of the errors from measurement of immobilized cells : To calculate the probability, we first require the distribut ion profiles of and . Thirty randomly picked fixed (therefore being immobilized) NIH3T3 fibroblasts were recorded every minute ( = 1 minute) for 1 hour to document the position variations generated from noise of the tracking system and tracking softwa re . For each of the 30 cells, the cell centroid and nucleus centroid positions were determined [ 51 ] to obtained the 61 set of measurements C j * (t i ) and N j * (t i ), fo r i = 0 and j j is to identify individual cells. Since the cells were fixed, both C j (t i ) and N j (t i ) should be also constant, where C j and N j denote the true
48 positions of fixed cell centroid j and fixed nucleus cent roid j respectively. In this case, and for realized values of errors e . Assume that the measurements are unbiased, i.e., the errors have mean 0 in both the x and y coordinates. Then, C j and N j can be estimated as the average of the measured positions and , respectively, denoting as and , accordingly. In addition, the coordinates of and can represent the realized measurement errors for the cell and nucleus, respectively. The total number of errors that obtained from the 61 measurements of the 30 immobilized cells are 1830. While it may be common practice to model th e error distributions of the x and y coordinates as two independent Gaussian curves with equal variances, we found that the errors do not necessarily satisfy this assumption. Further, the errors between the x and y coordinates possess a slight but statist ically significant correlation. This might due to: 1) The vibration of the microscope while changing the filters consistently along one direction, and 2) The consistent fluorescence photobleachin g effect propagating from the edge of the cell onward the cen ter of the cells. Therefore, a non parametric approach is adopted instead of modeling the errors using Gaussian distribution. We shall give an empirical probability mass function (PMF), with equal probabilities assigned to each of the 1830 (x,y) co ordinate pairs . Likewise, shall be given an empirical PMF, with equal probabilities on each of the 1830 (x,y) coordinate pairs . Computation of the probabilities of interest for a single c ell : The establishment of and distributions from the fixed cells allows us to accurately estimate P(CCD > NCD) value for randomly selected mobile (free) cells. Hence, we
49 repeated the data acquisition process using the same microscopic s etup but replaced fixed cells to live cells. Here, we assume that the distributions of and for the fixed cells can represent those for the live cells since the acquisition systems are the identical. Since P(A) = for any event A , the value of P(CCD > NCD) can be computed based on empirical PMFs, where the errors are independent to each other. Hence, = ( 3 6) Using the empirical PMFs described in the previous section, we would have , and each sum runs over the 1830 possible values of the 4 terms is not compu tationally feasible, we therefore apply a Monte Carlo technique to evaluate this quadruple sum: let N be a large integer, then the value of the probability can be approximated closely as follows, Set count = 0. Repeat for Draw a value for from the empirical PMF of (i.e. randomly select one of the 1830 error values) Draw a value for from the empirical PMF of Draw a value for from the empirical PMF of Draw a value for from the empirical PMF of
50 Set If D > 0 , add 1 to the value of count . The Monte Carlo approximation of the probability is then . Intuitively, each D represents a possible CCD NCD true value, based on a set of measurement errors. By repeating the procedure a large number of times, we thus account for the randomness in the generating of measurement errors an d obtain by substituting these back into equation ( 3 5 ): ( 3 7) Comparison across differen t cell cycle phases : The average for a population of cells can be obtained by averaging the individual n s ingle cell picking from a random sample in that population. In this application we are particularly interested in comparing the average among K populations of cells, where each population represents a different cell cycle phase, so that we can detect motility differences between phases. Thus, we must perform a pairwise comparison of means. Since there is no reason to expect the only a moderate sample size of cells per population, the assumptions of the classi cal two sample t test may be violated. Thus, our preferred method would be a permutation test procedure on the samples, which also provides a simple way to adjust for the multiple comparisons that are of interest. We set the significance threshold fo r the permutation p value such that the overall chance of one or more type I errors (false positive) is 0.05. For reference purposes, we also show the results of the classical
51 3.1.3 Materials and methods Cell culture and sample pre paration : NIH 3T3 fibroblasts were obtained from ATCC (American Type Culture Collection, Manassas, VA) and maintained in cell culture media, DMEM (Mediatech Inc., Manassas, VA), supplemented with 10 % fetal bovine serum (Hyclone Laboratories, Logan, UT), 1 % L glutamine (Mediatech) and 1 % penicillin streptomycin (Mediatech). Cell culture were passed every 3 days and kept in an 37Â°C incubator containing 10 % CO 2 . pEGFP plasmids (BD Biosciences) were introduced into NIH3T3 fibroblasts following standard proto col of transfection using lipofectamine to get the humongous fluorescence over cell body (Invitrogen, Carlsbad, CA). Approximately 24 hours after transfection , the cells were transferred onto the glass bottom dishes (World Precision Instrument, Sarasota, FL) , which were coated with 0.01 % poly L lysine (Sigma Aldrich, St Louis, MO) and fibronectin with a final concentration of 20 ÂµL /m L (BD Biosciences, Bedford, MA), at a cell loading concentration to avoid cell cell interactions in accordance with the sing le cell level. Cell cycle synchronization : Double thymidine block was used to arrest cells at late G 1 phase [ 52 ] . First, thymidine (S igma Aldrich) was applied to cell cultures at 2 mM for 12 hours and then washed out with sequential replacements of culture media: HBSS (Mediatech) briefly, DMEM without FBS for 15 minutes and DMEM with FBS for another 15 minutes. Afterward, cells were kep t in fresh culture media for 9 hours before the second 2 mM thymidine arrest for another 12 hours. Finally, cells were released from thymidine again using the same procedure to achieve synchronization.
52 Flow cytometry ( BD Biosciences ) then was used to exa mine the cell cycle phases after the cells were synchronized for different time course. DNA of the collected cells were labeled with propidium iodide (Abcam) and an aylized with LSR II flow cytometer (BD Biosciences) and the data were analyzed by FACSDiva ( BD Biosciences) to evaluate the percentage of test cell population in each phase. The time course with the maximum percentage of test cell population at a specific phase was selected as the right check point. It was determined that the cells were in the S phase, G 2 phase, recessive early G 1 phase, and recessive late G 1 phase at 2, 5, 9 and 15 hours right after release . Microscopy for live cell image acquisition: Live cell i mage acquisition was achieved under the an environmental chamber with a CO 2 suppleme ntary system (In Vitro Scientific, St. Louis, MO) at 37 Â°C . The nuclei were stained by Hoechst 33342 (Sigma Aldrich) at 20 Âµg/m L 10 minute prior to the image acquisition to prevent any potential interference for the cell cycle progression. A n imaging acqui sition system (TE 2000; Nikon, Melville, NY), equipped with a 20Ã— objective lens, a fluorescent light source (X Cite 120 PC; EXFO, Ontario, Canada) and a CCD camera (Cascade:1K; Roper Scientific, Tucson, AZ) was used to document cells and the correspondi ng nuclei at one minute time intervals for 1 hour period through two channel fluorescent microscopy (the green channel for cells and the blue channel for nuclei). The boundaries and centroids of cells and nucleus were determined using a custom made program in MATLAB platform (The MathWorks, Natick, MA) [ 51 ] .
53 3 .1.4 Results GFP labeled NIH3T3 cells were successfully synchronized to obtain the predominate population at different cell cycle phases. It took 0, 2, 5, and 9 hours for NIH3T3 fibroblasts to reach the late G 1 , S, G 2 , and successive early G 1 phase after initial synchronization (denoted as T0, T2, T5, T9, respectively). 22 to 28 random sampled cells and their corresponding nuclei at each cell cycle phase were monitored for 1 hour at one minute intervals for their coordinates. The calculated, using N = 10,000,000 replications in the Monte Carlo procedure. We chose this N based o n repeatedly calculating several individual p i replicates were enough to reproduce values within a margin of 0.001, which is sufficient numerical accuracy. Therefore, our estimates of the population (denoted as pop ) duri ng the time periods 0 1 hr, 2 3 hr, 5 6 hr, 9 10 hrs after early G 1 synchronization are 0.574, 0.648, 0.691, 0.617. For the permutation test procedure, the critical value of the t statistic was found to be 2.58, in order to limit the overall chance of one or more type I errors from the six pairwise comparisons to be 0.05. The significant differences are T0 vs T2 (t = 2.67), T0 vs T5 (t = 4.13) and T5 vs T9 (t = 2.66). We found that the same three pairwise differences were significant at the 95 % level un adjustments for the multiple comparisons. The significant higher pop of cells in G 2 phase in comparison to both early and late G 1 phase demonstrates the more protrusion (or contraction) dominated behaviors of cells in G 2 phase tha n that of cells in both the tested G 1 significantly from early to late G 1 phase but significantly from G 2 to early G 1 phase.
54 Moreover, cells in S phase exhibit significant mo re protrusion (or contraction) dominated behaviors than that of cells in late G 1 phase, indicating the decreased active migratory status as cells progress from late G 1 to S phase. To better understand if cell migratory behavior is time interval associated , this model was further applied on the identifying pop s at different time intervals. The locomotion of cell normally lasts for 5 to 10 minutes [ 50 ] . Therefore, more bias would be generated for describing cell locomotion events if the time intervals of observations are too long. However, at longer time intervals, the capacity of cell motility will be more directly revealed with less interference of system error. In our study, we investigated the estimates of probability of CCDs greater than NCDs with non overlay measurements at time intervals of 1min, 2mins, 5mins, 10mins, and 15mins, respectively. pop s at different time intervals and the standard deviation of the population are show n in Table 3 1. pop s during G 1 phases, regardless of time inte rvals, always exhibit relative smaller values comparing to S and G 2 phase. It is notable that po p of G 2 phase keeps exhibiting the largest value until time intervals increases to 15mins. pop of S phase at 15mins intervals jumped up to be the largest, demonstrating less long term migration ability of S phase. However, it is noticeable that as time interval increases, the standard deviation increases due to larger fluctuations of s l ed by the increased factoring measurement error . At longer time interval, the data collected is not enough to be significant, but the trends indicates that G 1 phase possess the largest migration capacity while S phase exhibits the lowest migration capacity. This can be further supported by the 1 hour displacement quantification.
55 3.1.5 Discussion Most of the current methods of cell migration quantification focus on ut ilizing a good amount of long (~10 hours) trajectories to obtain the overall cell migration capacity therefore minimize or even out the measurement bias [ 53 ] . This is due to the fact that a cell may remain stationary for a certain period of time before suddenly migrating without any re gular pattern. However, if more detailed characterization of migratory behavior is pursued, a proper observation interval of cell movement is under demand. While the approximation to the integral of interest in equation (2) is mathematically valid for any , the estimate becomes increasingly coarse and biased if is too large (say, > 10 min), as locomotion events may be missed entirely. On the other hand, the approximation will also fail to provide useful information if . In that case, the finite difference approximation is too noisy to reveal the true speed; the measured displacements will be almost entirely error and probability model will thus assign ~0.5 to each miniscule time interval, yielding no information. An ideal should be small, yet large enough to reasonably distinguish CCDs and NCDs from the noise. In our study, we chose = 1 min as an efficient time interval of observation for the purpose of collecting data experimentally. It also appears to be a reasonable choice to allow for distinguishing measurements from noise. Given coordinates at the beginning and end of an interval, let M denote the measured displacement, T denote the actual true displacement. Then approximately we have , where is the error measurement variance in one coordinate direction. This relation allows us to estimate the average true displacement from our collected data. For CCDs at = 1, our data showed that , while , giving .
56 This indi cates that nearly 80 % of the measurement value comes from the true movement, demonstrating that one minute interval is long enough to make real movement and errors distinguishable, while sensitive enough to catch the subcellular movement without losing to o much information .If a significantly higher frequency than 1frame/min is recorde d, other methods, such as Kalman filter [ 54 ] , to smooth out the path should be considered for extracting the real movement information. However, this will require setting assumptions about t he trajectory dynamics of single cells, which are not currently well established. At 1 min interval, NIH3T3 fibroblasts exhibited the highest frequency of protrusion (or retraction) dominant behavior at G 2 phase, followed by S phase. However, the more freq noticeable that when time interval increases till 15 mins, pop at S phase star ts to jump over G 2 phase, demonstrating its more protrusion (or contraction) dominant behavior in a longer term. This indicates pop at longer time interval could have more likelihood to reveal cell migration capacity considering the smallest migration capacity of cells in S phase in from quantification of 1 hour displacement (Supplementary Fig. 1b). Meanwhile, cells at G 1 phase always maintained the highest frequency of detachment dominant behavior, regardless of time intervals. Some other studies, usi ng hourly nucleus displacement as tools, indicated that Lymphocyte migration likely happens in G 1 phase rather than S and G 2 /M phase . Moreover, it is investigated that t he motility of L929, HeLa and BT4Cn glioma cells are greater in G 1 /S phases compared to G 2 /M phases [ 55 ] .
57 The concept of using fixed cells to characterize the measurement error could also be apply on other cells types with different migratory featur es, such as epithelial cells or amoeboid cells, as long as CCDs and NCDs of free cells can be captured using the same system as for fixed cells. In our study, our statistics of interest is the fraction of time where CCS is greater than the corresponding NC S for a selected time interval. However, the extensive application of this method for an error contaminated system can also be the precise quantification of the movement with enough sensitivity and accuracy. This is pursued under the conditions such as tim e sensitive drug treatment, where the long term tracking is not sensitive to provide enough information. This systematic study utilizing the discrepancy (relativity) between CCDs and NCDs can also provide an insight in the quantification of a system with r elative movement, such as Chemotaxis, which requires high sensitivity of cell morphology and migration change towards the change of chemical supply. 3.2 The Relationship between the Cell Cycle and Cell Migration d uring Interphase of NIH3T3 Fibroblasts 3.2. 1 Introduction It has been reported before that individual Rho GTPases can affect the progression of the cell cycle. The inhibition of RhoA or the overexpression of the negative dominant form of Rac1 or Cdc42 can block mitogen stimulated G 1 /S transition i n Swiss 3T3 fibroblasts [ 26 ] . Contrariwise, the introduction of individual Rho GTPases into quiescent Swiss 3T3 fibroblasts is sufficient to induce this transiti on [ 26 , 27 ] . Furthermore, RhoA is also in charge of maintaining the correct timing for cyclin D1 expression duri ng G 1 phase in h 5 3T3 cells [ 28 ] ; and the activation of either Cdc42 or Rac1 can promote anchorage independent cell cycle progression in mouse embryonic
58 fibroblas ts [ 56 ] . These studies indicate that individual Rho GTPases can actively affect the cyclin CDK complexes in the progression of the cell cycle. Conversely, prote ins responsible for the cell cycle progression are also known to affect the activities of Rho GTPases. The cyclin dependent kinases inhibitors, p21 Cip1 and p27 Kip1 , can bind to the cyclin CDK complexes to arrest cells in G 1 phase [ 57 ] . When thrombin triggers endothelial cells to undergo mitogenesis, p27 Kip1 is down regulated and its CDK inhibitory activity is disrupted through phosphorylation. This event causes p27 Kip1 e ither to be degraded inside the nucleus or exported to the cytoplasmic region, where p27 Kip1 binds and inhibits RhoA before being degraded [ 58 ] . Similarly, p21 Cip1 also regulates cyclin CDK complexes. Even though it is not clear whether p21 Cip1 can be exported from the nucleus to the cytoplasmic region when its cell cycle arrest activity is disrupted, p21 Cip1 is present in the cytoplasmic region, nevertheless, to inhibit Rho associated kinase (ROCK) during G 1 /S transition before it is degraded [ 58 ] . Both RhoA and ROCK are the upstream proteins of LIM kinase (LIMK), which is known to inhibit ADF/cofilin for its F actin severing and G actin sequestering activities that can depolymerize stress fibers into high concentrated monomeric actin pool in cytosol to facilitate cell migration [ 59 , 60 ] . Hence, p21 Cip1 and p27 Kip1 might possess dual roles [ 61 ] : On one hand, they maintain the cell cycle in G 1 phase when they are in the nucleus. On the other hand, they promote cell migration when they are in the cytoplasm. However, it is still unknown whether the inh ibitions of RhoA and ROCK through these two proteins are a general theme appearing in the migratory cells, such as fibroblasts, or not. The comprehensive study of the relationship between the cell cycle and cell migration from the cell cycle progression pe rspectives is challenging. Direct cell
59 migration assessment usually comes from single cell trajectories [ 62 ] . Yet, a cell may remain stationary for a certain time period and suddenly move without any regular pattern. Hence, cell migration studies usually require a good amount of long (~10 hours) trajectories to evenly cover the possible cell migration modes of a cell type without a bias. However, this approach generates a paradox in studying the cell cycle dependent migratio n the duration of the sampled cell trajectories must be long enough to comprehensively describe different cell migration modes, yet short enough to ensure the analysis remain in a certain cell cycle phase. Therefore, a new approach that can analyze cell migration but avoid these two contradictory requirements is on demand. In this study, a new approach has been developed to statistically compare the simultaneous cellular and nuclear displacements in short time intervals to decipher the cell migration mode s. Combined with precise cell synchronization, we explore the specific cell migration pattern in different cell cycle phases to provide an integrated view toward the molecular mechanism controlling the relationship between the cell cycle and cell migration in NIH 3T3 fibroblasts. 3.2.2 Materials and methods Cell culture and culture media : NIH 3T3 fibroblasts were obtained from ATCC (American Type Culture Collection, Manassas, VA) and maintained in DMEM (Mediatech Inc., Manassas, VA), supplemented with 10 % fetal bovine serum (Hyclone Laboratories, Logan, UT), 1 % L glutamine (Mediatech) and 1 % penicillin streptomycin (Mediatech). Cell cultures were maintained in a 10 % CO 2 incubator and passed every 3 days when the cell culture reached ~ 70 % confluence. Cell cycle synchronization: Double thymidine block was used to achieve cell cycle synchronization [ 52 ] . Since thymidine residual in the cell culture could prolong the
60 cell cycle arrest, we modified the thymidine washing procedure to sharp the synchronization and deliver the maximal cell population in a certain cell cycle phase. First, thymidine (Sigma Aldrich, St Louis, MO) was applied to cell cultures at 2 mM for 12 ho urs to arrest cell cycle at G 1 /S phase. Then the cells were released from cell cycle arrest by a series of culture media replacements: HBSS (Mediatech) briefly, DMEM without FBS for 15 min and DMEM with FBS for 15 min. Afterward, cells were kept in fresh c ulture media for 9 hours before the thymidine arrest and release was repeated. Then, the cells were synchronized at the end of G 1 phase. Quantit ative PCR to measure mRNA level: The RNeasy Mini Kit (Qiagen, Valencia, CA) was applied to isolate the total RN A from cell samples. Consequently, the iScript Advanced cDNA Synthesis Kit (Bio Rad, Hercules, CA) was used to reverse transcribe the isolated RNA to cDNA. Then, the quantities of the target cDNA were determined by quantitative PCR (qPCR) using CFX connect system (Bio Rad). qPCR reactions were conducted with SYBR supermix (Bio Rad). The quantities of GAPDH in different samples were assessed as the standard in different samples. Intron spanning primers for RhoA (Genbank accession number: JN971019.1), Rac1 (N M_009007.2) , Cdc42 (U37720.1) , cyclin A1 (BC120518.1), cyclin E1 (BC138662.1), p21 Cip1 (AB017818.1), p27 Kip1 (U10440.1), and GAPDH (GU214026.1) were designed using Primer3 and synthesized by Integrated DNA Technologies (Skokie, IL) as follows: RhoA sense: AAACAGGATTGGCGCTTTTG RhoA antisense: CACAAGATGAGGCACCCAGA Rac1 sense: AAGTGTGTGGTGGTGGGAGA
61 Rac1 antisense: GTCAAAGACGGTGGGGATGT Cdc42 sense: CTGGGGCATCTTCGTGTCTT Cdc42 antisense: AACCCCATACACACCCCAAA cyclin A1 se nse: AGTACCTGCCTTCACTCATTGCTG cyclin A1 antisense: TCTGGTGAAGGTCCACAAGACAAG cyclin E1 sense: GCCCTCTGACCATTGTGTCC cyclin E1 antisense: GCACCACTGATAACCTGAGACCT p21 Cip1 sense: AATCCTGGTGATGTCCGACC p21 Cip1 antisense: T CAAAGTTCCACCGTTCTCGGG p27 Kip1 sense: CAGCTTGCCCGAGTTCTA p27 Kip1 antisense: GGGGAACCGTCTGAAACA GAPDH sense: TCTCTGCTCCTCCCTGTTCC GAPDH antisense: GTTCACACCGACCTTCACCA Four measurements (2 experimental duplicate and 2 biolog ical duplicate) were performed for each protein, and relative cDNA levels were determined using the comparative Ct method [ 63 ] . Rho GTPases Pull d own Assay an d Western blotting : NIH 3T3 fibroblasts were seeded at 5 Ã— 10 4 cells/ mL over night to obtain cells in the single cell context, and then the cell culture was synchronized by thymidine. Afterward, synchronized cells were released and harvested at the designa ted time courses (see results and discussion section); the cells were homogenized to obtain total proteins at 1 Âµg/ ÂµL in the supernatant. Finally, the supernatant were subjected to RhoA and Rac1/Cdc42 pull down assays (Cytoskeleton), respectively, to colle ct the active form of the proteins.
62 The samples subjected to Western blotting were separated by 12 % SDS PAGE and transferred onto a PVDF membrane. Thereafter, individual antibody against to the probed protein was applied to the PVDF membrane. After horse radish peroxidase conjugated secondary antibody was against the primary antibody, the image was developed using standard protocol. Then, the blotting results were quantified using ImageJ software (NIH, Bethesda, MD). The antibodies applied in this study w ere monoclonal RhoA (Santa Cruz Biotech, Dallas, TX), Rac1 (Millipore, Billerica, MA), Cdc42 (BD biosciences, Bedford, MA) a nd ROCK1 (Abcam, San Francisco, CA) antibodies. Ce ll immunofluorescence staining: For microtubule staining, the fixation buffer was composed of 0.5 % glutaraldehyde, 0.8 % formaldehyde and 0.5 % Triton X 100 in PBS and the blocking buffer was composed of 1% paraformaldehyde and 1 % BSA in PBS. For actin and vinculin staining, the fixation buffer was composed of 4 % paraformaldehyde and 0.5 % Triton X 100 in PBS, and blocking buffer was composed of 2 % BSA and 0.02 % Triton X 100 in PBS. Cells on glass coverslips were subjected to fixation buffer and blocking buffer sequentially for 30 min each at room temperature. Then, cells were incub ated with 1:400 dilution of mouse raised monoclonal anti tubulin (Invitrogen, Carlsbad, CA) or anti vinculin (Invitrogen) overnight at 4Â°C. After 3 times washing with PBS, the cells were subjected to 1:1000 dilution of Alexa Fluor 488 conjugated anti mou se IgG ( Invitrogen) for 1 hour at room temperature. Actin and nuclear were stained by 1:40 dilution of Alexa Fluor 568 phalloidin (Sigma Aldrich) and 1:100 dilution of Hoechst 33342 (Sigma Aldrich), respectively, mixed in the secondary antibody solution.
63 The antibodies used to detect p21 Cip1 and p27 Kip1 in the immunostaining process were 1:200 dilution of polyclonal p27 (C 19) (Santa Cruz) and 1:100 dilution of polyclonal p21 (C 19) (Santa Cruz), and Alexa Fluor 488 conjugated anti rabbit IgG (Invitrogen) were the secondary antibody. The staining procedure was the same as the vinculin staining process. M icroscopy for image acquisition : A TE 2000 imaging acquisition system (Nikon, Melville, NY) equipped with a 20Ã— objective lens, an X Cite 120 PC fluoresce nt light source (EXFO, Ontario, Canada) and a Cascade:1K CCD camera (Roper Scientific, Tucson, AZ) were used for image acquisition. The acquisition setting was set at 300 ms exposure time, 3Ã—3 binning, and 25 % light intensity from the light source. C1 con focal laser scanning system (Nikon) with a 60x oil immersion objective (Nikon) was used to acquire the confocal images . Live cell imaging : Glass bottom dishes (World Precision Instrument, Sarasota, FL) were pre treated with 0.01 % poly L lysine (Sigma Ald rich) and 20 Âµl/ml fibronectin (BD Biosciences). A CO 2 supplementary system (In Vitro Scientific, St. Louis, MO) was operated during image acquisition to maintain the experimental environment at 10 % CO 2 and 37 Â°C. pEGFP plasmids (BD Biosciences) were intr oduced into NIH 3T3 fibroblasts using lipofectamine with standard protocol (Invitrogen). Right after transfection, the cells were passaged onto the glass bottom dishes at a concentration in the single cell level to avoid cell cell interactions. After 24 ho urs, green fluorescent protein (GFP) was produced to fluorescently label the cells. When nuclei were subjected to fluorescently label, Hoechst 33342 (Sigma Aldrich) was applied to the cell culture at 20 Âµg/ mL 10 min
64 right before image acquisition to preven t any potential interference to the progress of the cell cycle from the nuclear stain [ 64 ] . Then the cells were documented through two channel fluores cent microscopy (the green channel for cells and the blue channel for nuclei) at the demanding time intervals. Live cell image analysis and quantification : The cell boundaries were determined using a robust algorithm developed in our lab [ 51 ] using MATLAB platform (The MathWorks, Natick, MA) . A proper background intensity threshold was selected for each image frame to determine the boundary of nucleus appearing in t he image. Fixed fluorescent beads bombarded on the glass bottom slides were used to align images of the same stacks of images in a movie to minimize the positioning error generated from image acquisition systems [ 62 , 65 ] . Cell cycle curve fitting in nuclear intensity histogram : The nuclear intensity histogram was generated from ~ 5000 nuclei and plotted with 50 bins. This histogram was further fitted by the Dean Jett Fox model [ 66 ] to determine the percentile of cells in each cell cycle phases. Briefly, the G 1 and the G 2 /M peaks were individually fitted to a normal (Gaussian) distribution curve with th e same coefficient of variations using least squares approximation. Data in between the two normal distribution curves were fitted by the superposition of a broadened second order polynomial and a normal curve to obtain curve describing the distribution of cells in S phase. The area covering each phase was used to determine the percentiles of cells presenting in the phase of the cell cycle. Elimination o f systems error for CCD and NCD : Each centroid data contained errors introduced to the positioning measu rement by the acquisition system. However,
65 the system errors could not be identified from a single image. Instead, the errors can be estimated using the error profile elucidated by repeatedly acquiring the centroid of the same object from a fixed cell. Onc e the sample size n was large enough, the geometric distribution of the measured centroids could be treated as the error profile of the acquisition system. Then, the geometric center of the error profile was superposed onto the centroid of the object from a live cell to determine the distribution domain of the n would generate n possible true centroid positions surrounding the measured centroid after the superposition. Accordingly, the measured centroid had carried the weight of the error profile with it. centroids of the same object between two consequent stack images. Since in each image frame there were n troids, it would derive n x n displacements. Hence, the system errors could be reduced from the displacement by averaging the n x n possible displacements. The cell cycle pha se determined by flow cytometry : Cells were trypsinized, centrifu ged to form a pellet, and washed three times with ice cold PBS. Consequently, the pellet was fixed with 70 % ethanol overnight at 4Â°C . Next day, the pellet was re suspended in PBS containing 100 mL ribonuclease, and incubated at room temperature for 30 min. Afterward, the sample was subjected to 50 mL propidium iodide (Abcam) to label the DNA at room temperature for 30 min before the samples were analyzed with LSR II flow cytometer (BD Biosciences). The percentage of cells in each phase of the cell cycle was estimated using FACSDiva (BD Biosciences).
66 3.2.3. Results and Discussion NIH 3T3 fibroblasts change their dynamic patterns du ring th e cell cycle : NIH 3T3 fibroblasts were used as the model cell line to explore whether a relationship exists between the cell cycle and cell migration. First, we recorded the motions of GFP labeled NIH 3T3 fibroblasts for a whole cell cycle from one division to the next. The cell centroid displacements over the whole cycle were extracted at 3 min intervals (denoted as CCD 3 min ) and plotted against the corresponding time course. It was showed that the CCD 3 min values were greater in the first half of t he cell cycle and significantly dropped after the halfway ma rk (Fig. 3 2 ). H ence, the short term displacements of a cell might be related to the stage of its cell cycle. The cell cycle phases of the synchronized NIH 3T3 fibroblasts can be controlled : In o rder to study specific cell migratory behaviors under different cell cycle phases, successful synchronization of cell cycle is under demand. However, currently the cell cycle phases are mainly identified by flow cytometry, which requires the cells to be tr ypsinized, and therefore the adherent cell migration status cannot be probed. Since the DNA content in the nucleus defines the cell cycle phase, the nuclear DNA amount could be adopted to determine the cell cycle phase while cell migration was analyzed whe n the cells adhere on the substrate. Hence, Hoechst 33342 was used to fluorescently label the nuclei so that the nuclear intensity could reveal the amount of DNA in the nucleus. Consequently, the nuclear intensity histogram was constructed from ~ 5000 rand omly selected single cells (i.e., no cell cell interactions that inhibit cell cycle progression) and fitted to the Dean Jett Fox model to determine the nuclear intensity ranges for cells in each cell cycle phas e (Fig. 3 3 A ) [ 66 ] .
67 Foll owing, the time required for a culture of synchronized cells to propagate to a desired cell cycle phase after releasing from cell cycle arrest in the later G1 phase was determined by comparing the peak intensity of each nuclear intensity histogram (n ~ 500 0) generated from the synchronized cultures at different time frames to the peak intensities in the nuclear intensity histogram of the asynchronous cells, presenting individual cell cycle phases. Accordingly, the optimal time for the predominant population of synchronized cells to reach the late G 1 , S, G 2 , successive early G 1 , and successive late G 1 phase after initial synchronization were 0, 2, 5, 9, and 15 hours (denoted as t 0 , t 2 , t 5 , t 9 , and t 15 , respectively) ( Fig. 3 3 B ). The correspondence of the cell cycle phases to the propagating time for the synchronized NIH 3T3 fibroblasts was validated by flow cytometry (Fig. 3 3 C ) . Cell migration can be analyzed using the coupled nuclear and cellular displacements : Since cells can be synchronized into G 1 , S, G 2 and M phases, the cell motion changes could be studied in individual cell cycle phases. However, to comprehensively analyze the cell migration mode within a cell cycle phase is challenging. It is known that cell motion has an irregular pattern: a single c ell may remain stationary for an uncertain period of time, before suddenly beginning to move. Hence, a cell migration study usually requires a good amount of long (~10 hours) trajectories to evenly cover the possible cell migration modes of a cell type wit hout bias [ 62 ] . However, this approach generates a paradox in studying the cell cycle dependent migration the duration of the sampled cell trajectories must be long enough to comprehensively describe different cell migrat ion modes, yet short enough to ensure
68 the analysis remains within a certain cell cycle phase. Therefore, we need to bypass the long trajectories to study cell migration. An active migratory mesenchymal cell would alternatively protrude its leading edge an d detach its trailing edge to migrate. In contrast, a sampling cell only routinely performs membrane protrusions and retractions in random edges but its body (and nucleus) does not move. Hence, these two migration modes active migration and sampling ar e composed of two cell locomotion events: protrusion (or retraction) and detachment. A trailing edge membrane detachment event often gives rise to approximately the same degree of nuclear translocation. Yet, a centroid displacement is calculated by averagi ng the local area change by the overall area. Because the nucleus always moves as a whole, the nuclear centroid displacement (NCD) is equal to the quantity of nuclear translocation. On the other hand, the coupled membrane detachment only gives rise to a sm all portion of the cell area change, which is divided by the whole cell area to yield a smaller cell centroid displacement (CCD). Hence, a membrane detachment event impacts the NCD greater than the coupled CCD; i.e., NCD > CCD. In contrast, in a cell protr usion (or retraction) event the nucleus tempts to stay still (NCD ~ 0) while the CCD is non negligible, so CCD NCD. Together, the difference between short term CCD and the coupled NCD could provide insights into cell migration to distinguish these two lo comotion events. Even though the duration of a cell locomotion event varies, it generally lasts for 5 to 10 minutes. Hence, with longer time intervals (e.g., 15 min), more bias will be resulted from using the measurements alone to describe a cell locomotio n event. We need time intervals shorter than the duration of a cell locomotion event, so that the
69 event can be dissected into small fractions for analysis. Each small fraction of time can then be ascribed to a specific locomotion event; by accumulating the fractions, a weight describing the composition of the overall locomotion in terms of NCD > CCD and CCD > NCD events is obtained. However, the values of CCD and NCD become miniscule and indistinguishable from the measurement errors (determined by calibrati ng on immobilized cells) when the time intervals are too short (< 0.5 min). Therefore, we adopted a time interval of 1 min for this approach. Once enough short term NCD and coupled CCD data from randomly picked cell motion movies are collected, the weighti ng compositions of the NCD dominant (NCD > CCD) and CCD dominant (CCD > NCD) events can be used to describe the specific cell migration mode of a probed cell type. An active migratory cell should have relatively equal weights of NCD dominant and CCD domina nt events while a pure sampling cell should have significantly more CCD dominant events. Hence, the weight of the CCD dominant (or NCD dominant) events can be a useful index to quantitatively survey the cell migration mode. We call this analysis a CN discr epancy analysis. NIH 3T3 fibroblasts possess specific migration modes and migration capacities in each cell cycle phase : The CN discrepancy analysis was then applied to the movies of NIH 3T3 fibroblasts to evaluate their cell movement modes in each cell c ycle phase. GFP introduced NIH 3T3 fibroblasts were synchronized at the late G 1 phase (time t 0 ) and consequently labeled by nuclear dye, Hoechst 33342 . Then they were monitored at times t 0 , t 2 , t 5 , t 9 , and t 15 , which correspond to late G 1 phase, projected S phase, G 2 phase , successive e arly G 1 phase, and successive late G 1 phase, respectively. Afterwards , movies of 22 to 28 randomly selected single cells and the
70 coupled nuclei were recorded at one minute intervals for 1 hour to extract the (one minute) NCD s and coupled CCDs in each cell cycle phase. Following, the CCD dominant fraction s were calculated ( Fig. 3 4A ). The results showed that the cells in G 2 phase possessed the highest CCD dominant fraction ( F c = ~ 69 % , n = 25 ); followed by the S phase cells ( ~ 65 % , n = 28 ), the early G 1 phase cells (~ 62 % , n = 24 ), and finally the late G 1 phase cells (~ 57 % for t 0 , n = 22 and ~ 56 % for t 15 , n = 24 ). We would expect F c to be stable in the two late G 1 phases ( t 0 and t 15 ) , and they were indeed similar in our data. Applying Tukey's HSD to cross compare F c values of different phases , there were significant pairwise differences in F c values between these phas es at a 95 % confidence level: late G 1 and early G 1 , G 2 and late G 1 , G 2 and early G 1 . This analysis provid ed first hand information regarding the make up of cell locomotion in different cell cycle phases of NIH 3T3 fibroblasts. The results suggested that the movement modes are dependent on cell cycle phase, as evidenced by the differing F c values. In addition, the movement modes between the early and late G 1 phase or between S and G 2 phases possessed no significant different, although the CCD dominant fractions were different. However, th e CCD dominant fraction information needed to be further combined wit h the magnitudes of the nuclear displacements to determine the cell migration status. Thus, we probed the 1 hour nuclear trajectories of synchronized cells (n = 30) and found that the average 1 hour nuclear displacements were 10.86, 6.09, 8.58, 9.96, and 1 t 0 , t 2 , t 5 , t 9 , and t 15 test showed that the 1 hour displacements of S phase ( t 2 ) cells were significantly different from those of cells in other phases (Fig. 3 4B ).
71 Taken together, detailed cell moveme nt modes in different cell cycle phases can be elucidated. Since the average 1 hour nuclear displacement is highest and the F c value is lowest in later G 1 phase, fibroblasts migrate most actively in this period. Once cells enter S phase, the F c value is sl ightly greater than that in later G 1 phase but the average 1 hour nuclear displacement drops to its minimum. Hence, the cell bodies barely move and the cells possess a fair amount of sampling activity. In G 2 phase, the F c value increases to the greatest an d the average 1 hour nuclear displacement also increases from S phase but does not reach the level in G 1 phase, so cells have limited cell migration but undergo active sampling. After mitosis, cells cycle back to G 1 phase and regain their migratory capacit y until late G 1 phase as indicated by the F c values and the average 1 hour nuclear displacement values. It is also noteworthy that the cells have more protrusion (or retraction) activities in early G 1 than later G 1 phases based on the F c values. These conc lusions can be supported by the stack images, which showed that cells have more detachment events in G 1 phase than S and G 2 phases; also, the detachment events are not observed in S phase cells (Fig. 3 4 C ). The actin cytoskeleton development and focal adhe sion is cell cycle associated : Cell morphology, governed by the actin cytoskeleton, is highly associated fibroblasts to compare their morphology, actin structure and the di stribution of the cell substrate contacts in each cell cycle phase (Fig. 3 5 A ). The microscopic images revealed that in G 1 phase actin filaments were dispersed, fuzzy and could not be clearly defined in cells. Meanwhile, the cell substrate contact points were unevenly disseminated over the cell body as small patches, mainly
72 displayed on the perinuclear region and the cell edges. When cells entered S phase, the actin filaments developed into thick stress fibers over the whole cell body but more prominent in the cortical region. Accordingly, the cell substrate contact points became matured as focal adhesions, which were established into large, aligned patches along the stress fibers toward the cell edges. In G 2 phase, thick stress fibers remained the same ove r the entire cells. Interestingly, focal adhesion turnover had occurred, in which the size of the cell substrate contact points were not as significant as they were in S phase but more similar to those in G 1 phase. However, the smaller focal contacts still aligned in highly order with the stress fibers as they were in S phase but not in G 1 phase. Confocal microscopy was further applied to reveal the detailed structure within a thin focal plane (Fig. 3 5 B ). It supp orted the observation that the focal adhesio ns were fully developed in S phase and the sizes of cell substrate contact points presented by vinculin were relatively the same between late G 1 phase and G 2 phase. Taken together, the actin cytoskeletal patterns are unique in each cell cycle phases and in agreement with their specific cell migratory status. The Rho GTPases activities guide the morphological and migratory development throughout G 1 and S phases : Signaling pathways govern the action of detachment and protrusion (or retraction) , which are the results of actomyosin contraction and lamellipodia formation, respectively. Since the weights of these two locomotion events in each cell cycle phase are revealed by the CN discrepancy analysis, the contribution of the underlying signaling pathways also c an be specified in each cell cycle phase.
73 Rho GTPases direct actin cytoskeleton remodeling; so they must also play pivotal roles in receiving signals propagating from proteins involved in regulating cell cycle to alter the cytoskeletal organization and ce ll migration mode. Accordingly, the activity profiles of individual Rho GTPases in different cell cycle phase were assessed by the Rho GTPase activity assay s ( Fig. 3 6 A ). The results revealed that RhoA activity increased after cells passing G 1 /S transition , and then dropped significantly to the lowest level when the cell reached G 2 phase. Rac1 activity stayed at the same level in S phase as it was in G 1 phase, but drastically decreased in G 2 phase. In addition, Cdc42 activity continually descended through G 1 , S to G 2 phase. Our assessments also showed that RhoA activity in early G 1 phase was lower than that in later G 1 phase, whereas both Rac1 and Cdc42 activities in early G 1 phase were higher than those in late G 1 phase. We further probed the mRNA profiles of Rho GTPases and compared them with the activity profile s ( Fig. 3 6 B ). It wa s shown that the mRNA profiles and activity profile for each member of the Rho GTPases were consistent during the interphase. Therefore, the expressions of Rho GTPases should ref lect their activity demand closely. The activity evolution of Rho GTPases in G 1 and S phases can be explained smoothly: Once cells pass mitosis and re enter G 1 phase, they need to re position themselves from the highly specified cytoskeleton organization for cell division. Hence, nascent adhesions form to redistribute their substrate anchorages. Through the new focal contact formations, focal adhesion kinase activates Rac1 and Cdc42 and suppresses RhoA by inactivating p190RhoGAP [ 67 69 ] . However, after more focal adhesions become maturation in later G 1 phase, the FAK/paxillin complex enhances
74 RhoA activity again to induce actomyosin contraction and lowers Rac1 an d Cdc42 activities [ 70 ] . Therefore, the activities of Rac1 and Cdc42 decrease through early G 1 phase to late G 1 phase. Owing to that NIH 3T3 fibroblasts migrat e the most actively in later G 1 phase, one could consider that the profile of the three Rho GTPases is in their optimal for cell migration during this stage. The need to elevate the RhoA activity in S phase is supported by the fact that its mRNA amount is also elevated during the same period. The boost of RhoA activity from a normal migratory cell promotes the formation of stable stress fibers and focal adhesions in the cells to inhibit cell migration [ 11 , 71 ] . Hence, the increase of the RhoA activity, potentially only led by the maturation of focal adhesions, could explain the heavily increase of stress fibers in the cells and the significant decrease in cell motility after the cells enter S phase as we have observed in our assessments. Cytoplasmic p21 Cip1 and p27 Kip1 are closely associated with Rho GTPases and expression levels of ROCK throughout the cell cycle : When cells enter G 2 phase, the activities of all three Rho GTPases drop to the minimal levels as elucidated by the activity profiles of the Rho GTPases. Under this condition, Rho GTPases would not have the capacities to effectively alter the existing acti n cytoskeletal organization. This explains why the stress fibers, having been established during S phase, are preserved in G 2 phase as shown in the actin micrographs. It has been known that p21 Cip1 and p27 Kip1 are two prominent candidates to mediate the c rosstalk between the cyclin CDK complex and Rho GTPases [ 72 , 73 ] . T he functions of these two proteins are determine d by their cellular locations [ 74 ] . In the
75 cytoplasmic region, p21 Cip1 and p27 Kip1 affect the activities of RhoA and ROCK, respectively, to c hange the outcome of the actin cytoskeletal organization. Hence, to understand the role of ROCK in the process, NIH 3T3 fibroblasts were subjected to Western blotting against ROCK antibody in different cell cycle phases to assess the change in its expressi on level in different cell cycle phases. It was revealed that the expression level of ROCK was only slightly higher in S and G 2 phase than in G 1 phase (Fig. 3 7 A ). H ence, ROCK activity changes should be mainly governed by RhoA activity and the amount of cy toplasmic p21 Cip1 . To identify the expressions and locations of p21 Cip1 or p27 Kip1 in each cell cycle phase, NIH 3T3 fibroblasts were subjected to immunofluorescence microscopy against these two proteins. The results showed that these two proteins were pre sent inside the cells throughout the interphase (Fig. 3 7 B ). It was also revealed that the cytoplasmic concentration of p21 Cip1 was high in G 1 , decreased in S phase, but increase again in G 2 phase; and that of p27 Kip1 was none in G 1 , high in S phase, and l ow in G 2 phase. When this result was combined with the ROCK and RhoA concentration in each cell cycle phase, it was suggested that cytoplasmic p21 Cip1 and p27 Kip1 are indeed involved in the cell morphology and migration mode changes. Hence, besides the pr ofile of Rho GTPases discussed above, the impact of cytoplasmic p21 Cip1 and p27 Kip1 can also be incorporated to deliver the full picture of the cell activity relationship regarding the cell cycle and cell migration and their underlying molecular interactio ns. In G 1 phase, the presence of high concentration of cytoplasmic p21 Cip1 could inhibit the ROCK activity in cells regardless the status of RhoA activity; thereby increasing the ADF/cofilin activity for cells to undergo active migration. In S
76 phase, the h igh RhoA activity, high ROCK expression and low cytoplasmic p21 Cip1 concentration together lead the high ROCK activity in the cells, which promotes the developments of stress fibers and focal adhesions and results a low ADF/cofilin activity. Therefore, the cells exhibit dense stress fibers and are immobile. During the same period, the presence of cytoplasmic p27 Kip1 would not have an immediate impact to the high RhoA activity. However, the continuous increase of cytoplasmic p27 Kip1 due to the cell cycle pro gression [ 74 ] could gradually inhibit more and more RhoA, thereby steadily weakening the originally strong RhoA/ROCK activity throughout S ph ase. The decreases of RhoA and ROCK are evidenced by the occurrence of focal adhesion turnover after the cells passing S/G 2 transition, which requires a weaker RhoA activity but effective ROCK activity [ 75 78 ] . During the descent of RhoA/ROCK activities, the focal adhesion turnover proceeds. The continuously decrease of RhoA activity through the increasing of cytoplasmic p27 Kip1 eventually leads to the lowest RhoA act ivity in G 2 phase among the interphase. In G 2 phase, the shape and size of focal adhesions are much different from those in S phase but similar to those in G 1 phase, supporting the occurrence of focal adhesion turnover during S/G 2 transition. Once cells en ter G 2 phase, the re elevation of cytoplasmic p21 Cip1 strongly inhibit ROCK activity, thereby increasing the ADF/cofilin activity again. As a result, cells become more migratory in G 2 phase than S phase, yet the low activities of the Rho GTPases keep the c ells from active migration as in G 1 phase.
77 Table 3 1 . Statistics (estimates and standard deviations) of the fraction of time where CCS exceeds the corresponding NCS of NIH3T3 fibroblasts at different cell cycle phases (T0, T2, T5, T9) and at different ti me intervals ( s). Hours after synch Number of cells Statistics Interval (min) 1 2 5 10 15 0 22 pop 0.574 0.572 0.584 0.510 0.561 std dev 0.088 0.110 0.154 0.199 0.272 2 28 pop 0.648 0.662 0.674 0.658 0.665 std dev 0.073 0.098 0.166 0.202 0.217 5 25 pop 0.691 0.695 0.669 0.662 0.619 std dev 0.093 0.110 0.129 0.161 0.246 9 24 pop 0.617 0.594 0.532 0.507 0.443 std dev 0.100 0.120 0.163 0.239 0.232
78 Figure 3 1. Migratory behavior can be classified into two differen t categories: Protrusion (or contraction) dominant cellular behavior (left) and detachment dominant cellular behavior (right). A ) In protrusion (or contraction) dominant behavior, cell leading front moves in a larger scale with smaller (or no) nucleus tran slocation, translated into CCD NCD. In detachment dominant behavior, nucleus locates at the trailing end of the cell moves at a large scale of movement, translated into CCD
79 Figu re 3 2 . The displacements of NIH 3T3 fibroblasts exhibit a cell cycle dependent property. Just divided, green fluorescent protein (GFP) labeled single NIH 3T3 fibroblasts were monitored at 3 min time intervals over the whole cell cycle until next division. The cell centroid displacements at the same time intervals ( CCD 3 min ) were extracted to plot against time. The red and the blue curve individually represent the moving average of CCD 3 min of a monitored cell (moving step = 10).
81 Figure 3 3 . A) The nuclear intensity histogram (red) of asynchronous cells was fitted by the Dean Jett Fox model to segment each phase and determine its peak intensity: G 1 (orange curve), S (purple) and G 2 (green). B ) The optimal time frames to obtain the maximal population of synchronized cells in a desired cell cycle phase were determined by matching the peak intensity of a histogram generated from synchronized cells at a specific time frame after the cells were released from cell cycle arrest to the peak intensities gaine d from the histogram of asynchronous cells. The optimal time frame for synchronized cells propagating to late G 1 , S, G 2 , successive early G 1 and successive late G 1 phase are 0, 2 nd , 5 th , 9 th and 15 th hours (denoted as t 0 , t 2 , t 5 , t 9 , and t 15 ), respectively . C ) Flow cytometry revealed the stage of the cell cycle phases of synchronized NIH 3T3 fibroblasts at those time frames. The vertical dashed lines indicate the nuclear intensities of cells in G 1 , S, and G 2 /M phase of the cell cycle, respectively. All nucl ear intensity histograms were constructed with 50 bins (blue dots) based on ~ 5000 randomly selected nuclei, labeled by Hoechst 33342. D ) The mRNAs of cyclin A and cyclin E were quantified at t 0 , t 2 , t 5 , t 9 , and t 15 (left to right) to verify the cell cycle synchronization. Error bars represent the standard deviation (n=2).
82 Figure 3 4 . Cell migration modes are different in each cell cycle phase. A ) In each cell cycle phase, 25 one hour synchronized GFP labeled cell movies and their coupled nuclear movi es were subjected to the CN discrepancy analysis. The weights of the CCD dominant events in different cell cycle phases were showed at 1 min time interval . B ) The average one hour displacements were calculated for cells in later G 1 , S, G 2 , successive early G 1 and successive later G 1 , respectively. Double stars were denoted significantly different in the student T student test. Error bars represent the standard errors. C ) The stack images of GFP labeled cells in different cell cycle phases. Hollow and solid arrow indicated a protrusion (or retraction), and a detachment event, respectively.
83 Figure 3 5 . The organizations of actin c ytoskeleton and focal adhesions are associated with the cell cycle phases in NIH 3T3 fibroblasts. A ) The F actin stained rev ealed the actin cytoskeleton organizations in different cell cycle phases (top row); the vinculin stained revealed the distributions and sizes of cell contact points in each cell cycle phases (bottom row). B ) More detailed structures of actin cytoskeleton (top row) and vinculin stained focal contact points were showed by confocal images in different cell cycle phases. Scale bars: 3 0
84 Figure 3 6 . The activity and mRNA profiles of Rho GTPases vary in different cell cycle phases. A ) Western blotting results revealed the activity profile of Rho GTPases in different cell cycle phases after pull down assays (left) and the activity profile were quantified and normalized by the activity level of the t 0 (G 1 phase) samples (right). B ) The mRNA profile of Rho GTPases was tested in the synchronized cells of different cell cycle phases. In both panels, e rror bars represent the standard errors from two experiments.
85 Figure 3 7 . The expressions and localizations of both p27 Kip1 and p21 Cip1 changes in different cell cycle phases . A ) The expression levels of ROCK in different cell cycle phases were also assessed. Inset , the Western blotting result. Error bars represent the standard errors from three assays. B ) p27 Kip1 (left column), p21 Cip1 (right column) and nuclei of NIH 3T3 fibroblasts were fluorescently labeled to probe the expression and localization of p27 Kip1 and p21 Cip1 in NIH 3T3 fibroblasts. Scale bars: 100 .
86 Figure 3 8 . Schematic diagram of t he molecular interaction s mechanisms that lead the dynamics of focal adhesion and stress fiber formation s and their attribution to cell migration during cell cycle progression. The amounts of pu rple dots and green triangles in the S and G2 panels represent the quantity of p21Cip1 and p27Kip1, respectively (non proportional). The up and down arrows in the parentheses indicate the elevat ion and modulation of individual proteins activities , respect ively. Solid arrows indicate direction interactions and dash arrows indicate
87 C HAPTER 4 CONCLUSION AND FUTURE WORK The study on effect of LIG on T98G GBM provides the sight of using a highly sensitive tool for drug screening and revealed the potential effect of LIG on the treatment of GBM disease , which has barely been elucidated before . We demonstrated that the concentration of LIG that can efficiently hinder migration ability of T98G cells is as low as 5 ÂµM w ithout leading to the induction of apoptosis or reduction of cell proliferation rate . It was only when the concentration of LIG was raised to 2 5 ÂµM that LIG was found to be cytotoxic to T98G cells by arresting them at G 1 /S phase. However, Cell features, su ch as apoptosis and cell proliferation , which not provide as much sensitivity as single cell migration assay in this study, were the primary interest to investigate if a novel drug is applied traditionally . Due to the consistent result of quantifica tion of both collective cell migration and single cell migration, we were able to exclude the possible effect of cell cell interaction . Interestingly, we also did the same test on NIH 3T3 fibroblasts cell line. T he result indicated no significant change on either cell cycle arrest of migration capacity of NIH 3T3 cells before and after the treatment of LIG at the tested concentration (e.g. 5 ÂµM for migration assays, 25 ÂµM for cell cycle arrest test) . The presence/ absence of effect of LIG on cell migration and the cell cycle of both cancer cells and normal cell s spontaneous ly also implied the possible connection between cell cycle and cell migration . Additionally, the significant reduction of expression levels of th e three Rho GTPases, which are known to be both molecular switches for actin cytoskeleton remodeling and also key regulat ors for cytokinesis leads to the possible connection of cell cy cle and cell migration in the perspective of shared signal transductio n pathways.
88 To establish a framework between cell migration and the cell cycle regulation in a general case, we applied the integrated study of the cell cycle regulation on cell migration on a model cell line, NIH 3T3 fibroblasts , during inte rphase . We fi nd that fibroblasts maintain active migration state during G 1 phase. Once entering S phase, the cells become stationary with few protrusions (or retractions) . In G 2 phase, cell motility is predominant by sampling with few migratory events and the nuclear t ranslocation is significantly smaller than that of G 1 phase. Since in each cell cycle phase cells possess a unique migratory pattern and morphology, the cell phenotype may be treated as a subset of cell cycle phase dependent properties. Conversely, specifi c cell phenotypes may also serve as indicators to determine the cell cycle phase for the probed fibroblasts. The molecular interaction regulating this cell cycle associated cell migration was also investigated in our study. Even though some information of the molecular interactions has been explained in the literature, a proper and systematic explanation of the cell cycle cell migration association has t been established yet. We indicated the development of actin filament and focal adhesion along cell cyc le , which is responsible for the different types of subcellular behavior during different phases. This phenomenon was further found to be contributed by the co regulation of p21 Cip1 and p27 Kip 1 and small Rho GTPases. In fact, efforts has also been made to fit cellular behaviors under different cell cycle phases into a model from a different perspective , utilizing the distributions of microtubules location and actin stress fiber location, as indicated in (citation). This model can be subdivide d into 4 differ ent types: oriented non migrating, non polarized randomly oriented non migrating cell, randomly migrating, fully polarized
89 migrating . In is not hard to notice that the location of microtubule is also cell cycle associated (Fig. 4 ) . This elucidation involvi ng polarization regulation of cells during different cell cycle phases are one of the future studies we want to look into. The conclusion also helps to explain how cell proliferation and cell migration can work together in concert in organogenesis and wou nd healing process. When the proliferation rate is high, normal cells do not have a high migration activity since they have less time to stay in G 1 phase to actively migrate. In opposite, when normal cells have a lower proliferation rate but not in the qui escent state, they would have higher chance to stay in G 1 phase, thereby havi ng high migration activity. The correlation between parameters that could be extracted from these two scopes: cell cycle and cell migration , can also be of great interest worthwhi le studying , for instance, the correlation between duration of G 1 phase and cell proliferation rate s . Current existing cell migration analyses only reveal the final migratory status of the probed cells. Information generated from those analyses cannot dire ctly elucidate which protein and how a protein through signaling pathways alters the migratory outcome. In this study, we propose to use the CN discrepancy approach to reveal the cell migration status by quantitatively elucidating the weighing contribution s of different cell motilities to cell migration. The weights of different locomotion events can reveal the predominant molecular pathways in a specific migration pattern. There could also be potential development of this method. Utilizing CN discrepancy to study cell locomotion events actually provide possibility of precisely identifying angles between steps, and the actually step size. All this information can be closely related to any molecule of interest we want to look at. Once enough information
90 betw een these parameters and molecular regulation are collected, the molecular information can be explicated more straightforward from the cellular assessments . After entering the molecular biology era three decades ago, much effort has been devoted to shed l ight on the molecular mechanisms controlling cellular processes. Now, how different cellular processes work together to construct the cell physiology still needs to be c onnected. Since a protein in which the activity is shared by more than one cellular process needs to simultaneously satisfy these processes to ensure smooth cell physiology, the relationship disclosed in this study could be applied to functional genomics t o promote the understanding of cell physiology.
91 Figure 4 1 . Microtubule s distribution is cell cycle associated. The tubulin stain of both asynchronous and synchronized NIH3T3 fibroblasts revealed the microtubule distribution in different cell cycle phases. The bright green spots in A) and the red dots in B) indicate the location of centrosome s .
92 REFERENCES 1. Kurosaka S, Kashina A (2008) Cell biology of embryonic migration. Birth Defects Res C Embryo Today 84: 102 122. 2. Wolfe K (2003) Evolutionary biology: Speciation reversal. Nature 422: 25 26. 3. Friedl P, Gilmour D (2009) Collective cell mig ration in morphogenesis, regeneration and cancer. Nat Rev Mol Cell Biol 10: 445 457. 4. Bhatt A, Kaverina I, Otey C, Huttenlocher A (2002) Regulation of focal complex composition and disassembly by the calcium dependent protease calpain. J Cell Sci 115: 34 15 3425. 5. Schluter DK, Ramis Conde I, Chaplain MA (2012) Computational modeling of single cell migration: the leading role of extracellular matrix fibers. Biophys J 103: 1141 1151. 6. Nobes CD, Hawkins P, Stephens L, Hall A (1995) Activation of the small GTP binding proteins rho and rac by growth factor receptors. J Cell Sci 108 ( Pt 1): 225 233. 7. Nobes CD, Hall A (1995) Rho, rac and cdc42 GTPases: regulators of actin structures, cell adhesion and motility. Biochem Soc Trans 23: 456 459. 8. Nobes CD, Ha ll A (1995) Rho, rac, and cdc42 GTPases regulate the assembly of multimolecular focal complexes associated with actin stress fibers, lamellipodia, and filopodia. Cell 81: 53 62. 9. Kole TP, Tseng Y, Jiang I, Katz JL, Wirtz D (2005) Intracellular mechanics of migrating fibroblasts. Mol Biol Cell 16: 328 338. 10. Parri M, Chiarugi P (2010) Rac and Rho GTPases in cancer cell motility control. Cell Commun Signal 8: 23. 11. Ridley AJ, Comoglio PM, Hall A (1995) Regulation of scatter factor/hepatocyte growth fact or responses by Ras, Rac, and Rho in MDCK cells. Mol Cell Biol 15: 1110 1122. 12. Hanahan D, Weinberg RA (2000) The hallmarks of cancer. Cell 100: 57 70. 13. Onn A, Fidler IJ (2002) Metastatic potential of human neoplasms. In Vivo 16: 423 429. 14. Decaeste cker C, Debeir O, Van Ham P, Kiss R (2007) Can anti migratory drugs be screened in vitro? A review of 2D and 3D assays for the quantitative analysis of cell migration. Med Res Rev 27: 149 176.
93 15. Hayot C, Debeir O, Van Ham P, Van Damme M, Kiss R, et al. ( 2006) Characterization of the activities of actin affecting drugs on tumor cell migration. Toxicol Appl Pharmacol 211: 30 40. 16. Elledge SJ (1996) Cell cycle checkpoints: preventing an identity crisis. Science 274: 1664 1672. 17. Grana X, Reddy EP (1995) Cell cycle control in mammalian cells: role of cyclins, cyclin dependent kinases (CDKs), growth suppressor genes and cyclin dependent kinase inhibitors (CKIs). Oncogene 11: 211 219. 18. Johnson DG, Walker CL (1999) Cyclins and cell cycle checkpoints. Annu Rev Pharmacol Toxicol 39: 295 312. 19. Wajed SA, Laird PW, DeMeester TR (2001) DNA methylation: an alternative pathway to cancer. Ann Surg 234: 10 20. 20. Brady H, Desai S, Gayo Fung LM, Khammungkhune S, McKie JA, et al. (2002) Effects of SP500263, a novel , potent antiestrogen, on breast cancer cells and in xenograft models. Cancer Res 62: 1439 1442. 21. Buckley MF, Sweeney KJ, Hamilton JA, Sini RL, Manning DL, et al. (1993) Expression and amplification of cyclin genes in human breast cancer. Oncogene 8: 21 27 2133. 22. Locascio A, Nieto MA (2001) Cell movements during vertebrate development: integrated tissue behaviour versus individual cell migration. Curr Opin Genet Dev 11: 464 469. 23. Bonneton C, Sibarita JB, Thiery JP (1999) Relationship between cell mi gration and cell cycle during the initiation of epithelial to fibroblastoid transition. Cell Motil Cytoskeleton 43: 288 295. 24. Cho RJ, Huang M, Campbell MJ, Dong H, Steinmetz L, et al. (2001) Transcriptional regulation and function during the human cell cycle. Nat Genet 27: 48 54. 25. Tsai MA, Waugh RE, Keng PC (1996) Cell cycle dependence of HL 60 cell deformability. Biophys J 70: 2023 2029. 26. Yamamoto M, Marui N, Sakai T, Morii N, Kozaki S, et al. (1993) ADP ribosylation of the rhoA gene product by bo tulinum C3 exoenzyme causes Swiss 3T3 cells to accumulate in the G1 phase of the cell cycle. Oncogene 8: 1449 1455. 27. Olson MF, Ashworth A, Hall A (1995) An essential role for Rho, Rac, and Cdc42 GTPases in cell cycle progression through G1. Science 269: 1270 1272.
94 28. Welsh CF, Roovers K, Villanueva J, Liu Y, Schwartz MA, et al. (2001) Timing of cyclin D1 expression within G1 phase is controlled by Rho. Nat Cell Biol 3: 950 957. 29. de Curtis I, Meldolesi J (2012) Cell surface dynamics how Rho GTPases orchestrate the interplay between the plasma membrane and the cortical cytoskeleton. J Cell Sci 125: 4435 4444. 30. Chao WW, Lin BF (2011) Bioactivities of major constituents isolated from Angelica sinensis (Danggui). Chin Med 6: 29. 31. Liang MJ, He LC (2 006) Inhibitory effects of ligustilide and butylidenephthalide on bFGF stimulated proliferation of rat smooth muscle cells. Yao Xue Xue Bao 41: 161 165. 32. Liang MJ, He LC, Yang GD (2005) Screening, analysis and in vitro vasodilatation of effective compon ents from Ligusticum Chuanxiong. Life Sci 78: 128 133. 33. Yang J.Y. aCHH (2012) Advances in studies on pharmacological functions of ligustilide and their mechanisms. Chinese Herbal Medicines 4: 26 32. 34. Lu Q, Qiu TQ, Yang H (2006) Ligustilide inhibits v ascular smooth muscle cells proliferation. Eur J Pharmacol 542: 136 140. 35. Lin Q. ZAG, Chen J.N., Lai X.P., Gui S.H. (2011) Anti inflammatory and analgesic effects of ligustilide. Chin J Exp Tradit Med Form 17: 165 168. 36. Wang Y.N. DJR (2010) Studies o n cerebral ischemia in rats and anti inflammatory mechanism of ligustilide. West China J Pharm Sci 25: 685 687. 37. Zhu Z.G. ZRY (2004) The prevention and treatment of ischemic stroke. Mod Prac Med 16: 51 54. 38. Kuang X, Yao Y, Du JR, Liu YX, Wang CY, et al. (2006) Neuroprotective role of Z ligustilide against forebrain ischemic injury in ICR mice. Brain Res 1102: 145 153. 39. Tsai NM, Chen YL, Lee CC, Lin PC, Cheng YL, et al. (2006) The natural compound n butylidenephthalide derived from Angelica sinensis inhibits malignant brain tumor growth in vitro and in vivo. J Neurochem 99: 1251 1262. 40. Winklbauer R, Selchow A, Nagel M, Angres B (1992) Cell interaction and its role in mesoderm cell migration during Xenopus gastrulation. Dev Dyn 195: 290 302. 41. Qi H, Zhao J, Han Y, Lau AS, Rong J (2012) Z ligustilide potentiates the cytotoxicity of dopamine in rat dopaminergic PC12 cells. Neurotox Res 22: 345 354.
95 42. Quan Xia PZ, Shaoping Li, Yitao Wang (2004) Current study on Angelica sinennsis. Lishizhen Medicin e and Materia Medica Research 15: 164 165. 43. Wu G, Shi L, Hu J, Li L (1998) [The determination of ligustilide in Ligusticum chuanxiong Hort. by supercritical fluid extraction]. Yao Xue Xue Bao 33: 457 460. 44. Vial E, Sahai E, Marshall CJ (2003) ERK MAPK signaling coordinately regulates activity of Rac1 and RhoA for tumor cell motility. Cancer Cell 4: 67 79. 45. Petrulio PYMaC (2008) Activation of Rho Kinase by TNF activation in human pulmonary microvascular endothelial cells. The Jo urnal of Immunology 180: 550 558. 46. Qi H, Han Y, Rong J (2012) Potential roles of PI3K/Akt and Nrf2 Keap1 pathways in regulating hormesis of Z ligustilide in PC12 cells against oxygen and glucose deprivation. Neuropharmacology 62: 1659 1670. 47. Sheetz MP, Felsenfeld DP, Galbraith CG (1998) Cell migration: regulation of force on extracellular matrix integrin complexes. Trends Cell Biol 8: 51 54. 48. Kim DH, Cho S, Wirtz D (2014) Tight coupling between nucleus and cell migration through the perinuclear ac tin cap. J Cell Sci 127: 2528 2541. 49. Carlsson AE, Sept D (2008) Mathematical modeling of cell migration. Methods Cell Biol 84: 911 937. 50. Dong C, Slattery MJ, Rank BM, You J (2002) In vitro characterization and micromechanics of tumor cell chemotactic protrusion, locomotion, and extravasation. Ann Biomed Eng 30: 344 355. 51. Arce SH, Wu PH, Tseng Y (2013) Fast and accurate automated cell boundary determination for fluorescence microscopy. Sci Rep 3: 2266. 52. Jackman J, O'Connor PM (2001) Methods for s ynchronizing cells at specific stages of the cell cycle. Curr Protoc Cell Biol Chapter 8: Unit 8 3. 53. Kipper MJ, Kleinman HK, Wang FW (2007) New method for modeling connective tissue cell migration: improved accuracy on motility parameters. Biophys J 93: 1797 1808. 54. Wu PH, Agarwal A, Hess H, Khargonekar PP, Tseng Y (2010) Analysis of video based microscopic particle trajectories using Kalman filtering. Biophys J 98: 2822 2830. 55. Walmod PS, Hartmann Petersen R, Prag S, Lepekhin EL, Ropke C, et al. (20 04) Cell cycle dependent regulation of cell motility and determination of the role of Rac1. Exp Cell Res 295: 407 420.
96 56. Philips A, Roux P, Coulon V, Bellanger JM, Vie A, et al. (2000) Differential effect of Rac and Cdc42 on p38 kinase activity and cell cycle progression of nonadherent primary mouse fibroblasts. J Biol Chem 275: 5911 5917. 57. Chu IM, Hengst L, Slingerland JM (2008) The Cdk inhibitor p27 in human cancer: prognostic potential and relevance to anticancer therapy. Nat Rev Cancer 8: 253 267. 58. Konecny FA (2010) Review of cellular and molecular pathways linking thrombosis and innate immune system during sepsis. J Res Med Sci 15: 348 358. 59. Zhang S, Tang Q, Xu F, Xue Y, Zhen Z, et al. (2009) RhoA regulates G1 S progression of gastric cancer cells by modulation of multiple INK4 family tumor suppressors. Mol Cancer Res 7: 570 580. 60. Ohashi K, Nagata K, Maekawa M, Ishizaki T, Narumiya S, et al. (2000) Rho associated kinase ROCK activates LIM kinase 1 by phosphorylation at threonine 508 within the activation loop. J Biol Chem 275: 3577 3582. 61. Munoz Alonso MJ, Acosta JC, Richard C, Delgado MD, Sedivy J, et al. (2005) p21Cip1 and p27Kip1 induce distinct cell cycle effects and differentiation programs in myeloid leukemia cells. J Biol Chem 280: 18120 18129. 62. Wu PH, Arce SH, Burney PR, Tseng Y (2009) A novel approach to high accuracy of video based microrheology. Biophys J 96: 5103 5111. 63. Schmittgen TD, Livak KJ (2008) Analyzing real time PCR data by the comparative C(T) method. Nat Protoc 3 : 1101 1108. 64. Siemann DW, Keng PC (1986) Cell cycle specific toxicity of the Hoechst 33342 stain in untreated or irradiated murine tumor cells. Cancer Res 46: 3556 3559. 65. Wu PH, Nelson N, Tseng Y (2010) A general method for improving spatial resoluti on by optimization of electron multiplication in CCD imaging. Opt Express 18: 5199 5212. 66. Fox MH (1980) A model for the computer analysis of synchronous DNA distributions obtained by flow cytometry. Cytometry 1: 71 77. 67. Miyamoto S, Akiyama SK, Yamada KM (1995) Synergistic roles for receptor occupancy and aggregation in integrin transmembrane function. Science 267: 883 885. 68. Arthur WT, Burridge K (2001) RhoA inactivation by p190RhoGAP regulates cell spreading and migration by promoting membrane prot rusion and polarity. Mol Biol Cell 12: 2711 2720.
97 69. Bass MD, Morgan MR, Roach KA, Settleman J, Goryachev AB, et al. (2008) p190RhoGAP is the convergence point of adhesion signals from alpha 5 beta 1 integrin and syndecan 4. J Cell Biol 181: 1013 1026. 70 . Grashoff C, Hoffman BD, Brenner MD, Zhou R, Parsons M, et al. (2010) Measuring mechanical tension across vinculin reveals regulation of focal adhesion dynamics. Nature 466: 263 266. 71. Takaishi K, Sasaki T, Kato M, Yamochi W, Kuroda S, et al. (1994) Inv olvement of Rho p21 small GTP binding protein and its regulator in the HGF induced cell motility. Oncogene 9: 273 279. 72. Croft DR, Olson MF (2006) The Rho GTPase effector ROCK regulates cyclin A, cyclin D1, and p27Kip1 levels by distinct mechanisms. Mol Cell Biol 26: 4612 4627. 73. Lee S, Helfman DM (2004) Cytoplasmic p21Cip1 is involved in Ras induced inhibition of the ROCK/LIMK/cofilin pathway. J Biol Chem 279: 1885 1891. 74. Besson A, Assoian RK, Roberts JM (2004) Regulation of the cytoskeleton: an onc ogenic function for CDK inhibitors? Nat Rev Cancer 4: 948 955. 75. Ren XD, Kiosses WB, Sieg DJ, Otey CA, Schlaepfer DD, et al. (2000) Focal adhesion kinase suppresses Rho activity to promote focal adhesion turnover. J Cell Sci 113 ( Pt 20): 3673 3678. 76. Zhu X, Ohtsubo M, Bohmer RM, Roberts JM, Assoian RK (1996) Adhesion dependent cell cycle progression linked to the expression of cyclin D1, activation of cyclin E cdk2, and phosphorylation of the retinoblastoma protein. J Cell Biol 133: 391 403. 77. Balles trem C, Hinz B, Imhof BA, Wehrle Haller B (2001) Marching at the front and dragging behind: differential alphaVbeta3 integrin turnover regulates focal adhesion behavior. J Cell Biol 155: 1319 1332. 78. Kole TP, Tseng Y, Huang L, Katz JL, Wirtz D (2004) Rho kinase regulates the intracellular micromechanical response of adherent cells to rho activation. Mol Biol Cell 15: 3475 3484.
98 BIOGRAPHICAL SKETCH Jun received her bachelor s degree in Pharmaceutical Engineering from Xi an University of Technology, an d had her internship at Pharmaceutical Institution of Hong Kong Polytechnic Universit y working on pharmacokinetics of live animals and pharmaceutics. She then came to University of Florida and completed her m aster s d egree in Chemical Engineering in 2011 f ocusing on biological wastewater treatment with supervisory of Dr. Spyros Svoronos and Dr. Ben Koopman in a pleasant research environment . That effort was rewarded with a completed thesis , which then turned into one of her first author publications (also h er first corresponding author paper) . When she pursued her m aster s d egree, s he also worked on test of food quality in Deibel Lab oratory as an intern. She learned how to better organize herself and think independently with the help from her work . It is at the same year that she was recruited as a Ph. D. student in C h emical Engineering co advised by Dr. Svoronos and Dr. Yiider Tseng , and official ly started to become a member of Dr. Tseng s lab. She had her first author paper published in 2013, following by pa ssing her qualifying exam in the same year. She remained active in participating in conferences and research social activities. S he has been to 4 conferences till the end of 2014, and won best poster presentation on Pruitt Research Day in late 2013. Righ t after her first publication, she found the importance of the weight of statistics in her current study; therefore she decided to pursue her minor in statistics with support from h er advisors . In 2013, she developed a new project right after her first pub lication with the right guidance of Dr. Tseng and Dr. Svoronos , and then started to work on a project focusing on statistical methodology with the guidance from Dr.
99 Samuel Wong from Department of Statistics . She was able to develop 2 potential first author publications on both biology scope and statistics scope with the help from them. Right after she finished constructing both of the manuscripts to be submitted, she received her Ph.D. from the University of Florida in the spring of 2015. During these years in the lab, Jun Yin managed the group of approximately 5 m aster s students and 4 u nde rgrad students. Some of them had independent projects, and some of them helped with her projects . They eventually turned out to be co authors in her publication or potent ial publication. She maintained good relationship with everyone she worked with. She was also a big fan of sports throughout her education. She was a yoga coach and played badminton a lot. She also liked communication, and was curious about all the fresh t hings around her. She enjoyed the balance she worked on between her research life and personal life and hoped to keep it up and make it better in the future.