1 INVESTIGATIONS INTO THE BIOLOGY OF THE EXTRACELLULAR MATRIX IN NEUROGENESIS, BRAIN INJURY, AND CANCER By DANIEL JOSEPH SILVER A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILL MENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2012
2 2012 Daniel Joseph Silver
3 To my beautiful wife and darling daughter, Melinda and Olivia
4 ACKNOWLEDGMENTS None of th is work could have been accomplished without the help and support of a great number of people, to which I am eternally grateful. Firstly, Melinda, my wonderful wife of seven years and mother to our beautiful daughter Olivia, I cannot thank you enough. I am confident that without your love, your infinite patience, and support throughout my entire time as a graduate student, I could never have succeeded here. You and Olivia have given me the unquestioning love and support that I have so desperately needed You have been a source of motivation and a reason to strive for excellence when I would rather quit. I love you both so much. To my advisor and colleague Dennis. You have provided a tremendously rich environment to pursue my science, the freedom to succeed on my own, and the patience and guidance to help me manage my failures. Thank you for championing my efforts these many years. Also, I will never be able to fully thank you and Janice for the help and support that you gave to Melinda and I during pregnancy. We felt cared for as an extended member of your family during a truly dark and frightening time in our lives. To my dissertation committee, Drs. Reynolds, Scott, and Borchelt. I appreciate all of the advice and guidance that you have offered throughout my entire program here. Your generosity with your time and criticism has improved my work and my scientific understanding. Thank you. To all the members of the Steindler laboratory. Thank you all for the treme ndous support and attention you have given to my work and me. You have each unflinchingly offered your time and energy to help me design and execute experiments and made my
5 time in the lab a pleasure. You have all become a second family to Melinda and I. I have come to treasure each of the relationships that we share. To my friend and colleague Florian. We work so well together. My work, my time in the lab, and my life in general is better having come to know you. Without the countless hours of conv ersation, support, advice, and humor that you have given me, I know I could never have found success here. You and Dorit have become treasured members of my family. I cannot thank you enough and am looking forward to many more years of collaboration and friendship. To my brilliant student Michela. Your efforts pervade this manuscript. Your work was always excellent and given willing. It has been my absolute pleasure getting to know and work with you. Best of luck for an exceptionally bright future. To my wonderful parents, Mom and Marc. You have been an ever present source of support and encouragement. Your love and generosity has seen Melinda and I through these many years and we are forever grateful. I love you. To my Dad. I cannot thank yo u enough for the support and guidance that you have given me. You were the first person to champion my efforts as a scientist and have continued to challenge and guide me. I have learned so much from you. I am proud to follow in your footsteps and look forward to many more years of work together.
6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 9 LIST OF FIGURES ................................ ................................ ................................ ........ 10 LIST OF ABBREVIATIONS ................................ ................................ ........................... 12 ABSTRACT ................................ ................................ ................................ ................... 14 CHAPTER 1 BACKGROUND ................................ ................................ ................................ ...... 16 Identity of the Neural Stem Cell ................................ ................................ .............. 18 The Astrocytic Family ................................ ................................ .............................. 24 Neural Stem Cell Characteristics ................................ ................................ ............ 27 The Relationship Between Boundary and Neurogenic Astrocytes .......................... 28 Reactive Neur ogenic Astrocytes Following Injury or Disease ................................ 34 Abnormal Astrocytic Neurogenesis Following Oncogenic Transformation .............. 37 Glioma I nvasion and the Extracellular Matrix ................................ .......................... 44 2 CHONDROITIN SULFATE PROTEOGLYCANS POTENTLY INHIBIT INVASION AND SERVE AS A CENTRAL ORGANIZER OF THE BRAIN TUMOR MICROENVIRONMENT ................................ ................................ ........... 58 Results ................................ ................................ ................................ .................... 59 CSPGs Discriminate Between Invasive and Non Invasive Lesions .................. 59 Specific CSPG Core Proteins Define the Non Invasive Tumor ECM ............... 60 Reactive Astrocyte Responses Differentiate Invasive from Non Invasive Lesions ................................ ................................ ................................ .......... 61 Microglial Responses Differentiate Between Invasive from Non Invasive Lesions ................................ ................................ ................................ .......... 63 Reducing CSPG Mediated Inhibition Facilitates Brain Tumor Invasion ............ 64 Expression of LAR Phosphatase Receptor Distinguishes Non Invasive from Invasive Brain Lesions ................................ ................................ .................. 67 Addition of CSPGs Helps to Restrict Diffuse Infiltration ................................ .... 68 MMP Profile of Invasive and Non Invasive Lesions ................................ .......... 71 CSPGs Dicriminate Between Invasive and Non Invasive Human Tumors ....... 72 Methods ................................ ................................ ................................ .................. 73 Cell Culture ................................ ................................ ................................ ....... 73 Primary human cancer cell culture ................................ ............................. 73 Classic human cancer cell culture ................................ .............................. 75 Primary murine astrocyte cell culture ................................ ......................... 76
7 Primary murine microglial cell culture ................................ ........................ 77 Genetic Modification of Cancer Cell Lines ................................ ........................ 77 Aggrecan Laminin Spot Gradient Assay ................................ ........................ 78 Microglial Activation Assay ................................ ................................ ............... 80 In Vitro Tumor Dispersal Assay ................................ ................................ ........ 82 Intracranial Transplantation ................................ ................................ .............. 83 Surgical procedures ................................ ................................ ................... 83 Tissue preparation and immunohistochemistry ................................ .......... 83 Quantification of in vivo tumor invasion ................................ ...................... 84 Gelatin Substrate Zymography ................................ ................................ ......... 85 Human Tis sue Collection and Processing ................................ ........................ 86 Statistical Analysis ................................ ................................ ............................ 86 3 ZEB1 MEDIATES INVASION AND CHEMORESISTANCE OF GLIOBLASTOMA ................................ ................................ ................................ .... 95 Results ................................ ................................ ................................ .................... 95 Isolation and Characterization of Invasive Tumor Initiating Cells ..................... 96 Knockdown of ZEB1 Results in Decreased Invasion and Increased Chemo Sensitivity ................................ ................................ ................................ ...... 97 Mechanism of ZEB1 Mediated Chemoresistance ................................ ............. 98 The Complementary Nature of Invasion and Chemoresistance ....................... 99 Methods ................................ ................................ ................................ ................ 102 Cell Culture ................................ ................................ ................................ ..... 102 CFSE Loading and FACS ................................ ................................ ............... 102 Cell Viability Assay ................................ ................................ ......................... 103 Knockdown Experiments ................................ ................................ ................ 103 Animal Experiments ................................ ................................ ....................... 103 Immunohistochemistry and Immunocytochemistry ................................ ......... 104 Imag e Acquisition and Data Analysis ................................ ............................. 105 RNA Isolation and Quantitative Real time PCR ................................ .............. 105 Protein Isolation and Western Blotting ................................ ........................... 106 Statistical Testing ................................ ................................ ........................... 106 4 SEGREGATION OF HUMAN BRAIN TUMOR INITIATING CELLS AND GENES 118 Results ................................ ................................ ................................ .................. 119 Identification of Pertinent Brain Tumor Samples ................................ ............ 119 Isolation and Expansion of Heterogeneous Target Cell Populations .............. 120 Clonal Propagation and Comparative Analysis of Putative Target Cells ........ 122 Segregation of Tumorigenic ity Genes ................................ ............................ 123 Methods ................................ ................................ ................................ ................ 125 Cell Culture ................................ ................................ ................................ ..... 125 Immunocytochemistry. ................................ ................................ .................... 128 Xenograft Experiments. ................................ ................................ .................. 129 Molecular Biology. ................................ ................................ .......................... 129
8 5 SYNTHESIS ................................ ................................ ................................ ......... 140 LIST OF REFERENCES ................................ ................................ ............................. 151 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 162
9 LIST OF TABLES Table page 1 1 Neural Stem Like Astrocytes ................................ ................................ .............. 56
10 LIST OF FIGURES Figure page 1 1 An astrocytic neural stem cell derived neurosphere ................................ ........... 48 1 2 The astrocytic identity of the neural stem cell ................................ ..................... 49 1 3 Stem cell characteristics ................................ ................................ ..................... 50 1 4 Cartoon depicting the four examples of astrocytic stem cells ............................. 51 1 5 Tenascin C and CSPGs in the neurogenic subventricular zone ......................... 52 1 6 Boundary astrocytes are associated with developmentally regulated ECM proteins ................................ ................................ ................................ ............... 53 1 7 Injury associated reactive astrocytes up regulate developmentally regulated ECM proteins ................................ ................................ ................................ ...... 54 1 8 Human gliomas express reactive a nd neurogenic astrocytic cell markers .......... 55 2 1 CSPGs discriminate between diffusely invasive and non invasive brain lesions ................................ ................................ ................................ ................ 87 2 2 Reactive astrocytes respond differentially to invasive and non invasive lesions ................................ ................................ ................................ ................ 8 8 2 3 Microglial activation differs markedly between invasive and non invasive brain lesions ................................ ................................ ................................ ....... 89 2 4 Reducing CSPG mediated inhibition facilitates brain tumor invasion ................. 90 2 5 The CSPG receptor LAR differentiates invasive from non invasive lesions ....... 91 2 6 Co transplantation of invasive and non invasive tumors ................................ .... 92 2 7 CSPG expression inversely correlates with invasion in human clinical specimens ................................ ................................ ................................ .......... 93 2 8 Putative model of CSPG mediated invasion inhibition ................................ ........ 94 3 1 Isolation and characterization of invasive tumor initiating cells ........................ 107 3 2 Knockdown of ZEB1 results in decreased invasion and increased chemosensitivit y ................................ ................................ ............................... 108 3 3 Mechanism of ZEB1 mediated chemoresistance ................................ .............. 109
11 3 4 ZEB1 expression correlates with invasion in gliomas ................................ ....... 110 3 5 CFSE loading and FACS paradigm ................................ ................................ .. 111 3 6 Fluorescence micr ographs and threshold images of representative tumors ..... 112 3 7 F luorescence micrographs of N cadherin and ZEB1 ................................ ........ 113 3 8 The c omplementary nature of invasion and chemoresistance .......................... 114 3 9 Gene regulation of c MYB and ROBO1 ................................ ............................ 115 3 10 Full images of all western blots presented in Figure 3 3 ................................ ... 116 3 11 Full images of all western blots presented in Figure 3 4 ................................ ... 117 4 1 Glioma derived sphere forming cells from the neurosphere assay ................... 132 4 2 Cell derivation protocol ................................ ................................ ..................... 133 4 3 Distinguishable glioma cell populations ................................ ............................ 134 4 4 Explant cultures generate target cell populations ................................ ............. 135 4 5 hGBM L19 cells at early stages post engraftment ................................ ............ 136 4 6 Generation of cell clones from hGBM L19 cultures ................................ .......... 137 4 7 Preliminary studies on interdependent cell clones from a single glioma ........... 138 4 8 Segregation of tumorigeniciy genes ................................ ................................ 139
12 LIST OF ABBREVIATION S BSA Bovine Serum Albumin CFSE Carboxyflurescein Succinimi dyl Ester Chondroitinase ABC CNS Central Nervous System CSC Cancer Stem Cell CS GAG Chondroitin Sulfate Glycosaminoglycan CSPG Chondroitin Sulfate Proteoglycan CTGF Connective Tissue Growth Factor EC Ependymal Cell ECM Extracellular Matrix EGFR Epidermal Growth Factor Receptor EMT Epithelial Mesenchymal Transition FCS Fetal Calf Serum GBM Glioblastoma Multiforme GFAP Glial Fibrillary Acidic Protein HABD Hyaluronic Acid Binding Domain HSC Hematopoietic Stem Cell HSPG Heparin Sulfate Proteoglycan i PSC Induced Pluripotent Stem Cell LAR Leukocyte Antigen Related Phosphatase Receptor LOH Loss of Heterozygosity MGMT O 6 Methylguanine DNA Methyltransferase MMP Matrix Metalloproteinase NOD/SCID Non Obese Diabetic/Severe Combined Immunodeficiency
13 NSC Neura l Stem Cell OPC Oligodendrocyte Precursor Cell PXA Pleomorphic Xanthoastrocytoma RMS Rostral Migratory Stream ROBO1 Roundabout Homologue 1 SEZ Subependymal Zone SGZ Subgranular Zone SNP Single Nucleotide Polymorphism SVZ Subventricular Zone TMZ T emozolomid e WFA Wisteria floribunda agglutinin ZEB1 Zinc Finger E Box Binding Homeobox 1
14 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 Ph ilosophy INVESTIGATIONS INTO THE BIOLOGY OF THE EXTRACELLULAR MATRIX IN NEUROGENESIS, BRAIN INJURY, AND CANCER By Daniel Joseph Silver May 2012 Chair: Dennis A Steindler Cochair: Brent A Reynolds Major: Medical Sciences Neuroscience Despite tremendo us effort, glioblastoma multiforme (GBM) remains the most pervasive and lethal of all brain malignancies. One factor that contributes to this exceedingly poor clinical prognosis is the highly invasive character of the tumor. Not a focused mass of cells, GBM is characterized by the microscopic infiltration of tumor cells throughout the otherwise healthy brain. Contrariwise, non neural metastases to the brain, as well as select lower grade gliomas can be relatively more treatable, specifically because they do not invade. Instead, these tumors develop as self contained and clearly delineated lesions in many respects, distinct and separate from the surrounding brain. With this dissertation, we first present evidence that the fundamental switch between the se two distinct tumor pathologies invasion and non invasion is mediated extrinsically, through the tumor extracellular matrix (ECM). Specifically, well circumscribed, focused brain lesions are associated with a rich, heavily glycosylated, chondroitin sulfate proteoglycan (CSPG) containing, tumor ECM, whereas no detectable CSPGs are associated with diffusely infiltrative brain tumors. Secondly, we extend our work to the intrinsic regulators of invasion and demonstrate that a
15 subpopulation of highly inv asive and chemoresistant glioblastoma cells is maintained by the transcription factor ZEB1 (zinc finger E box binding homeobox 1). ZEB1 is preferentially expressed in invasive glioblastoma cells, and its knockdown results in a dramatic reduction of tumor i nvasion as well as increased sensitivity to the chemotherapeutic agent Temozolomide (Temodar, TMZ) in vitro and in vivo Lastly, we present findings on the cellular identities of the stem like tumor cells themselves that compose a single GBM. We examine the concept that GBM may not be a single disease, but rather the culmination of multiple, discrete stem cell pathologies that ultimately manifest as GBM. In all, this work represents several unique attempts to better understand the complex biology of hum an brain cancer.
16 CHAPTER 1 BACKGROUND Most contemporary neuroscientists and neurosurgeons were educated according to the classical view that the central nervous system (CNS) is a peculiarly non healing tissue. Severed nerve fibers generally do not regr ow to form appropriate synaptic connections, and certainly no new neurons are generated after perinatal development is completed. This view is colorfully encapsulated by the oft cited statement of Santiago Ramon y Cajal: (1928) Once development was ended, the fonts of growth and regeneration of the axons and dendrites dried up irrevocably. In the adult centers, the nerve paths are something fixed, and immutable: everything may die, nothing may be regenerated. It is for the science of the future to change, if possible, this harsh decree. The science of the future, many believe, has arrived in the form of the relatively recently discovered indigenous neural stem cells (NSC) that persist within the brain throughout life. While there was a small historical b ody of anti canonical reports hinting at persistent adult neurogenesis, appreciation for the existence of NSC within the mature brain became widespread only after the publication of innovative methods to culture and expand these cells under highly speciali zed conditions in the form of clonal neurospheres (see Figure 1 1). Since then, there has accumulated substantial literature demonstrating both the presence of multipotent NSC across virtually all regions of the neuraxis, and persistent and functionally r elevant neurogenesis within the subependymal zone (SEZ) and the hippocampal dentate gyrus in rodents and primates, including Man (reviewed in Taupin and Gage, 2002). In particular, the SEZ has been shown to generate mitotic neuronal precursors that underg o long range migration within the rostral migratory stream (RMS) to the olfactory bulb (Luskin, 1993). While the
17 majority dies en route to or within the olfactory bulb (Morshead and van der Kooy, 1992), many survive to differentiate and functionally integr ate as either granule cell or periglomerular cell interneurons that help modulate incoming olfactory sensory information from the nasal epithelium. Similarly, the subgranular zone (SGZ) of the dentate gyrus has been shown to contain mitotically active cel ls that give rise to neuroblasts that migrate out into the granule cell layer before differentiating and sending an axonal projection to the CA3 region (Seri et al., 2001; van Praag et al., 2002). Hippocampal neurogenesis is, of course, particularly germa ne to human neurological function, as it has been shown to be critically involved in learning and memory. Additionally, hippocampal neurogenesis has proven to be sensitive to a variety of modulating stimuli both positive and negative that can dramaticall y influence cognition (Kempermann et al., 1997; van Praag et al., 1999), suggesting that it may be possible to therapeutically augment this system to combat the memory decline that accompanies a variety of neurological diseases and insults. The connection between NSC activity and higher order brain function exemplified in the hippocampus, represents the root of the greatest hope associated with NSC; their potential use in cell replacement therapies. At some future point, such replacement may be accomplish ed either by in vitro expansion and transplantation, or by directing the migration of indigenous NSC pools towards areas of damage or cell loss. Achieving these future goals requires a deep understanding of the unique biology and activity of these cells, beginning with the most fundamental inquiries into their particular identity. Investigations into the identity of adult NSC over the past decade have established with high confidence that a special type of astrocyte is primarily responsible for both the
18 persistent neurogenesis seen in vivo in the subependymal zone and hippocampus, and the formation of multipotent neurosphere clones seen in vitro. While there is still no foolproof method for prospectively identifying which astrocytes in the CNS have NSC a ttributes, the identification of the cell type responsible for these phenomena opens the door for detailed study of their intrinsic capacity for self renewal and differentiation, as well as their ultimate suitability for therapeutic transplantation approac hes and their potential role in the formation and pathogenesis of neoplasias. In this review, we will present a historical narrative of some of the early studies that identified the astrocyte as a likely NSC candidate. Next, we will address the largely s emantic issue of whether the more appropriate. Then we will review the defining characteristics of NSC and compare these properties to those applied to the grandfather of all tissue specific stem cells, the hematopoietic stem cell (HSC). Finally, we conclude with a discussion of the possible role of astrocytic stem cells in the genesis and progression of brain tumors. Identity of the Neural Stem Cell An enduring roadbl ock to the study of NSC biology is the lack of effective prospective markers leading to isolation strategies from either whole tissues or cell cultures. The original descriptions of techniques for culturing NSC relied on a retrospective approach. That is a population of single cells was obtained from brain and cultured under optimizing conditions. Eventually, a percentage of these cells formed multipotent clones, proving in hindsight that the starting population harbored NSC. To prove self renewal, thi s process was typically repeated with single cell dissociates of these multipotent clones, and the formation of secondary multipotent clones was again retrospective evidence of the presence of NSC. While this approach
19 has been fruitful for localizing, in b road terms, the existence of NSC, prospective identification is needed if one wishes to understand and ultimately control the proliferation and differentiation of NSC and their progeny. For the sake of comparison, it is instructive to consider the HSC. O nce recognized only retrospectively by their ability to reconstitute the bone marrow of myeloablated hosts, living HSC can now be prospectively identified on the basis of the selective expression of a battery of surface antigens. In fact, this method of pr ospective isolation now allows for the highly efficient transplantation of single HSC that are capable of fully reconstituting ablated bone marrow (Osawa et al., 1996). As was true in the early search for the identity of the HSC, the initial investigations into the identity of the NSC began not by interrogating candidate NSC with antibodies against surface antigens, but rather by fractionating and testing subpopulations of previously identified cells within the system. Of cells within the mammalian CNS, th e immediately subjacent to the neurogenic SEZ, and seem, therefore, appropriately positioned to serve as NSC. More importantly, perhaps, EC have a phylogenetic history of f unctioning as NSC (Bruni, 1998). There is persuasive evidence that adult EC respond to injury in both reptiles and amphibians by dividing to generate replacement neurons throughout the CNS (see below for references). Moreover, EC have long been suspected of being responsible for the sporadic reports of adult mammalian neurogenesis following injury (Altman, 1962). Experimental evidence supporting the NSC role of EC in mammalian CNS was provided in one of the first functional tests of the NSC attributes of a prospectively identified neural cell type. Johansson and
20 colleagues (1999) examined the ability of EC to form multipotent neurospheres in vitro by pre labeling them via diI infusion into the lateral ventricle. Upon dissociation and sorting, they then s howed that labeled cells were capable of generating neurospheres with multilineage differentiation potential (i.e. neurons, astrocytes, and oligodendrocytes). Additionally, in support of their hypothesis that EC are the in vivo source of NSC, they showed that EC ringing the central canal of the spinal cord of adult rats became proliferative and generated astrocytes after spinal cord injury. These were very provocative findings, and caused quite a stir among the nascent NSC biology community. However, thi s paper was followed in rapid succession by a series of reports that immediately called into question the conclusion that EC are the NSC. Chiasson and colleagues, also in 1999, used mechanical microdissection to isolate the EC layer and showed that, while EC were capable of forming clonal structures, these structures were not true neurospheres since they were invariably unipotent, giving rise only to cells identified as astrocytes on the basis of GFAP antigenicity. Almost simultaneously, Alvarez Buylla an d colleagues repeated the diI labeling protocol used in the Johanssen study. Results from this study indicated that, while intraventricular diI does label the EC layer, the dye is apparently transferred to other cells either in vivo, or in vitro during the neurosphere culturing step (Doetsch et al., 1999). Additional experiments reported in this same study, using cellular tracers that do not show evidence of intercellular transfer (i.e. rhodamine beads, and adenovirus), revealed that EC labeled via a contr alateral ventricular injection of tracer failed to form neurospheres, and in fact disappeared from the cultures altogether. Neurospheres were obtained only from periventricular tissue obtained ipsilateral to the injection, and these injections invariably resulted in some
21 labeling of SEZ cells due to leakage of tracer from the needle penetration. Shortly thereafter, our laboratory published a study examining the ability of single dissociated EC, identified visually by their rhythmically beating cilia, to g enerate neurospheres (Laywell et al., 2000). Our results were concordant with those described in the Chiasson study individual EC formed only unipotent clones of GFAP+ astrocytic cells. These negative findings, combined with the lack of persuasive confir matory studies even to this day supporting the notion that EC have NSC attributes, strongly suggest that EC neither normally divide in vivo, nor generate multipotent neurospheres in vitro, and therefore are not likely to represent the NSC. These dispara te results are attributable possibly to discrete transfer of diI from EC to cells of the SEZ either prior to or during the dissociation step of neurosphere culture. It is still not known, however, why Chiasson and we were able to show unipotent clone form ation by cultured EC, while Doetsch using similar culture techniques was not. reduced to the population of cells that inhabit the SEZ the area of the greatest persistent neuroge nesis and the highest density of neurosphere forming cells (Weiss et al., 1996). Extensive ultrastructural analysis of the SEZ by the Alvarez Buylla laboratory identified three major constituents of this region (Doetsch et al., 1997). The most abundant S EZ cells are the highly proliferative neuroblasts, termed type A cells. Type A cells are the neuronally committed precursors that migrate through the RMS to the olfactory bulb where a fraction of them differentiate into either granule or periglomerular int erneurons. There is also a second highly proliferative cell with ambiguous phenotype, termed the type C cell, which seems to function as a transit amplifying intermediary
22 between the NSC and the migrating neuroblasts. Finally, there is a slowly dividing SE Z astrocyte, termed the type B cell, which seems to display many characteristic of the NSC. In addition to expressing GFAP, B cells were shown by ultrastructure to have irregular and invaginated nuclei, and their cytoplasm contained intermediate filaments and dense bodies. Subsequent experiments showed that the highly proliferative SEZ constituents can be ablated either by antimitotic drugs or ionizing radiation, but that the normal SEZ anatomy can regenerate after such depletion (Doetsch et al., 1999b; Ma rshall et al., 2005). By careful ultrastructural examination of the SEZ during ablation/regeneration, the Alvarez Buylla group (Doetsch et al., 1999b) showed that the B cell astrocyte survives the ablation, and divides to generate the C cell. The C cell in turn divides to generate the rapidly dividing type A cells that migrate to the olfactory bulb. This, then, suggested that a special type of astrocyte residing within the SEZ might be the NSC responsible for both persistent neurogenesis in vivo, and the generation of multipotent neurospheres in vitro. This same group showed that cells expressing GFAP can give rise to neurons in vivo. Using a transgenic mouse that allows for the selective retroviral infection of GFAP expressing cells with a constitutive reporter gene, these investigators showed that labeled astrocytes within the SEZ and SGZ give rise to olfactory bulb interneurons and hippocampal granule neurons, respectively (Doetsch et al., 1999a; Seri et al., 2001). At about the same time, we used the same transgenic system to show that astrocytes cultured from the adult mouse SEZ could form neurospheres capable of both glial and neuronal differentiation (Laywell et al., 2000). We also showed in this study that a subset of GFAP expressing astrocytic ce lls obtained from cerebral cortex, cerebellum, and spinal cord are capable of forming
23 neurospheres, but only when obtained from animals younger than about postnatal day 12. Since this age corresponds closely to the disappearance of radial glia in most reg ions of the mouse CNS, and since the neurogenic B cell astrocyte of Alvarez Buylla maintains some radial glia like morphological characteristics, we interpreted this latter result as evidence that radial glia or immature astrocytes are the in vivo represen tation of the NSC. Since these initial studies, there have been numerous confirmatory studies supporting the astrocytic identity of the NSC, at least in the adult, including human (Quinones Hinojosa et al., 2006) brain. That is not to say, however, that cells with astrocyte phenotype are the only cells with apparent attributes of multipotent progenitors. Kondo and Raff (2000) published a study showing that cultured oligodendrocyte precursors (OPC) isolated from the early postnatal rat optic nerve could b e induced to generate all three neural lineages by manipulation of the in vitro conditions, including the substrate and the presence of mitogens. These precursors were also able to form sphere like structures when cultured under conditions of anchorage wi thdrawal, but the potency of these structures was not reported. One unusual and interesting aspect of this study is that the conversion of OPC to multipotent progenitor was observed only after the OPC were induced to differentiate into type 2 astrocytes b y culturing in, and then withdrawing, serum, PDGF and BMP. Subsequently, another group reported that human white matter contains a glial progenitor cell, which can be prospectively enriched on the basis of CNP expression a protein associated with oligode ndroglial progenitor cells and can be induced to generate neurons and multipotent neurospheres (Nunes et al., 2003). It may in fact be that, with enough experimental manipulation, many different types of cells can be
24 induced to acquire NSC properties. Th ere is evidence that seemingly differentiated neurons can switch their phenotype to acquire astrocyte characteristics, including antigenic profile and membrane physiology (Okano Uchida et al., 2004; Laywell, et al., 2005). And we have proposed elsewhere t hat cells exist developmentally along a continuum stretching from non committed and multipotent to fully differentiated and functionally mature, but that there may be instances in which cell fate may move early ontogeny (Steindler and Laywell, 2003). Therefore, it is perhaps not surprising that a variety of cells show NSC attributes in vitro. Nevertheless, a preponderance of the evidence suggests that the cell responsible for normal maintenance of the per sistently neurogenic regions of the brain the NSC is a cell with apparent astrocytic phenotype. The Astrocytic Family Astrocytes that display adult NSC characteristics have persuasively been show to express the astrocytic cytoskeletal protein, GFAP (Imur a et al., 2003; Garcia et al., 2004). However, these NSC also lack many of the classically accepted characteristics of astrocytes, particularly functional features such as regulation of extracellular neurotransmitter concentration. Thus, there is a growing chorus of researchers who are uncomfortable with the designation of these cells as astrocytes, and agree with the s true that allowing diverse and non overlapping criteria to define a variety of astrocytic cells is unique for neural lineages. For instance, while there are many different subtypes of neurons, they can all be said to share the functional property of syn aptic transmission. Similarly, a cardinal feature of oligodendrocytes is the production of myelin, and the ensheathment of axons. So has
25 the NSC been inappropriately identified as an astrocyte? Should we reconsider this classification, and perhaps devel op new nomenclature to uniquely describe stem cells in the brain? We think not. For while we agree with Kimelberg (2003) that astrocytic maintain that it is approp riate to expand the functional definition of the astrocyte family to include the subpopulation of cells that possess NSC characteristics for at least three reasons: phylogeny, ontogeny, and convention. Phylogenetic analysis allows us to address the sometim es confusing phenotype of functionally related to the ependymoglial cell of lower vertebrates (see Figure 1 2). In lizards and reptiles, the amazingly multifunctiona l ependymoglial cell is the predominant astrocytic cell type (Reichenbach and Robinson, 1994; Bruni, 1998). Ependymoglia are radially oriented throughout life, and are thought to guide the migration of neuroblasts generated near the periventricular germin al matrix. Additionally, these cells have been shown to divide to generate functional neurons following lesions of both spinal cord (Egar and Singer, 1972; Nordlander and Singer, 1978; Anderson and Waxman, 1985; Simpson and Duffy, 1994), and cerebral cort ex (Molowny et al., 1995; Prez Caellas and Garca Verdugo, 1996; Font et al., 1997). Ependymoglia maintain ciliated connections with the ventricle, and are also responsible for phagocytic, injury response functions. Thus, it is clear that the functions encapsulated by the ependymoglial cell of the mammalian brain. Radial glia serve as migration scaffolds for newly generated neurons during early development, and hav e themselves been shown to be neurogenic
26 (Gaiano et al., 2000; Hartfuss et al., 2001; Noctor et al., 2002). Ependymal cells in the mammalian brain line and project cilia and microvilli into the ventricles (Spassky et al., 2005). Mature astrocytes can beco 2004). Finally, as discussed above, a subpopulation of astrocytic cells within the secondary germinal matrices of the SEZ and hippocampus are persistently neurogenic throughout life. Ontogenic analysis, too, reveals the interrelatedness of the major NSC candidates antecedents of stellate astrocy tes (Levitt and Rakic, 1980; Misson et al., 1991), and new evidence suggests that ependymal cells also develop from radial glia (Spassky et al., 2005). Expression of GFAP, at least during certain developmental stages or after injury, is also a feature sha red by radial glia, astrocytes, and ependymal cells (Takahashi et al., 2003), and there is evidence that mouse ependymal cells of the choroid plexus can differentiate into GFAP expressing astrocytes following transplantation to the injured spinal cord (Kit ada et al., 2001). Additionally, we and others have shown that clones of GFAP+ cells developed from prospectively identified EC (Chiasson et al., 1999; Laywell et al., 2000). Finally, convention suggests that we lump adult NSC features in with the multifu nctionality of astrocyte, rather than split these cells into a separate category. The fact is that, prima facie, adult NSC in vitro and in vivo simply seem like astrocytes. Yes, yes, a picker of nits can and will point out this or that astrocyte feature that is missing from astrocytic NSC, but there has for years been a large umbrella over the diverse
27 a pall of confusion over the field. Type I and II astrocytes are clearly distinct morphologically and functionally, and yet both are considered to be astrocytes. Likewise, Bergmann glia of the cerebellum and Mller glia of the retina are also commonly subsumed under the astrocyte designation. We believe that the over all phenotypic gestalt of the adult NSC is that of an astrocyte, and for the sake of common understanding we should refer to it as such. Neural Stem Cell Characteristics In comparison to other tissue specific stem cells, such as the HSC, the defining aspec ts of NSC are shrouded in ambiguity. This ambiguity can cause considerable confusion, and usually results in NSC biologist talking past researchers from other stem cell fields who generally have a much more rigorous definition of what constitutes a true st em cell (see Figure 1 3). In the initial glow of excitement over the descriptions of substantial adult neurogenesis and the persistence of neural stem like cells, the actual fundamental characteristics of NSC were, for the most part, poorly delineated, an d varied greatly between laboratories. For instance, within the persistent germinal matrices in vivo the SEZ and hippocampal dentate gyrus the overwhelming majority of stem cell progeny are neurons. Nevertheless, multilineage differentiation in the form of neurons, astrocytes, and oligodendrocytes is considered an indispensable in vitro characteristic of NSC. Self renewal is also generally required for NSC designation, although there is considerable confusion regarding how extensive self renewal should be in order to distinguish true NSC from long term progenitor cells. Most investigators would accept that one round of self But how many are required? By contrasting with the hematopoietic system, one can
28 intralaboratory variability in the antigenic profile used to select HSC from whole blood or bone marrow, there is general agreement that the HSC is defined function ally as: an extensively self renewing cell capable of multilineage differentiation, that can fully reconstitute depleted bone marrow in a serial transplantation paradigm. This means that a single HSC transplanted into a myeloablated host must reconstitute all of the blood isolate a single HSC from such a recipient and use it to reconstitute the blood lineages of a second myeloablated host for life. Clearly, then, fr om the perspective of the hematopoietic field, the salient attributes of NSC may seem relatively inadequate and poorly defined. Nevertheless, while recognizing the need to clearly define the in vitro and in vivo properties that characterize NSC, we believe that each tissue specific adult stem cell may require its own unique functional definition, and it may be inappropriate to force a simulacrum of the HSC definition onto the NSC. Indeed, to do so is to flirt with unnecessary scrupulosity given that the CN S is far more static that the hematopoietic system, and the constituent cells generally persist throughout life and are not replaced in toto. The Relationship Between Boundary and Neurogenic Astrocytes The subgranular zone (SGZ) of the hippocampal dentat e gyrus and the subventricular zone, (SVZ) lining the lateral walls of the lateral ventricles, represent the only two regions within the adult mammalian brain that support ongoing neurogenesis throughout life. (Figure 1 4) Within these rare germinal nich es, astrocytes, functioning as neural stem cells (Imura et al., 2003; Laywell et al., 2000), begin the cascade of events that continually renew the granule and periglomerular interneurons of the
29 olfactory bulb (Doetsch et al., 1999), and granule neurons of the adult hippocampus (Seri et al., 2001; van Praag et al., 2002). In vivo, the neurogenic process within the SVZ begins with a GFAP (glial fibrillary acidic protein) expressing astrocytic stem cell, or enic astrocytes, or what we refer to as multipotent astrocytic stem cells (MASCs)(Laywell et al., 2000; Chiasson et al., 1999; Garcia et al., 2004; Zheng et al., 2006) are the direct descendants of embryonic radial glia (Merkle et al., 2004; Merkle et al., 2007) and unlike mature cortical astrocytes, maintain a thin process, including a cilium, tethering them to the lateral ventricular wall and ventricular cavity (Doetsch et al., 1997). The relatively rare and quiescent B cells have been reported to give r ise to a population of highly proliferative, however less potent progenitor cells, referred to as transit amplifiers or C cells. Finally, from this putative intermediate progenitor pool, comes a population of young, immature neurons, or A cells that migra te forward over a tremendous distance through the rostral migratory stream (RMS) before a select few mature and integrate into the olfactory bulb neural circuitry as new interneurons (Doetsch et al., 1999; Kishi, 1987; Petreanu and Alvarez Buylla, 2002). Interestingly, a recent publication by Danilov and colleagues has multipotent B cell astrocyte may actually give rise to its neuroblast progeny directly, without the C cell interm ediate (Danilov et al., 2009). Coincidently, these results are in accord with our own in vitro model of SVZ neurogenesis, wherein cultured astrocytic stem cells are seen to robustly and directly generate a population of young neurons (Scheffler et al., 20 05). Whether the B cell astrocyte is directly neurogenic or requires a less potent intermediary, its activity is limited to the specific regions within the adult
30 brain where the expression of the extracellular matrix (ECM) is greatest. Under normal circu mstances, the ECM molecules CSPG, HSPG, and tenascin C are all intensely expressed strictly within the SVZ in the adult brain. (Figure 1 4) Thus, the coincidence of enriched matrix and persistent cell genesis, although still largely un quantified, must a ssuredly be of fundamental importance (Alvarez Buylla and Lim, 2004; Kerever et al., 2007). In the developing neonatal brain, neurogenesis is not quite so limited, yet it is still defined by a germinal ECM. Over 20 years ago, our laboratory and others d escribed transient patterns of enriched ECM expression within the early postnatal rodent brain. These matrix patterns appeared to outline functionally different emerging brain structures and, as such, engendered significant controversy over their particul ar role during brain development (for review see Steindler, 1993). During the course of these investigations, a subpopulation of astrocytes was brought to light that were consistently found intimately associated with these patterns. Situated within the E CM boundary itself, these unique cells became known as boundary associated astrocytes, or simply, 4A and 1 6) Although the original finding characterized these astrocytes within the developing somatosensory cortical barr el field (Cooper and Steindler, 1986a; Cooper and Steindler, 1986b), (Figures 1 4A and 1 6) subsequent investigations ultimately revealed boundary astrocytes associated with patterning neostriatal striosomes and subcortical brainstem nuclei (Steindler et a l., 1988), emerging olfactory glomeruli (Gonzalez Mde et al., 1993), the laminating optic tectum (Miskevich, 1999), as well as the roof plate of the embryonic spinal cord (Snow et al., 1990), suggesting that these interesting cells might be present extensi vely across
31 the entire neuraxis. At the time, we were unsure of the significance of these cells. With their characteristic hypertrophic GFAP expressing cytoskeleton and accompanied expression of the neurite growth inhibitory molecules CSPG and tenascin C the boundary astrocyte intimated a strong resemblance to the so of the adult brain. Interestingly, as we will elaborate below only now, in light of our current understanding of cytogenic astrocytes, have we begun to appreci ate that the behavioral and molecular hallmarks of neural stem cells may be present in the boundary astrocyte population. At first glance, boundary astrocytes may seem a bit out of place in the context of a stem cell story. They are certainly not, as ye t, members of the neural stem cell canon. After all, although these unique cells express many of the same immunophenotypic markers commonly associated with neural stem and progenitor cells, they have not yet been isolated ex vivo and tested for their abil ity to repeatedly self renew and generate clonal multipotent progeny. We believe that boundary astrocytes are potential stem cells: cells that typically do not function as stem cells but are capable of doing so under certain conditions (Potten and Loeffle following pieces of correlative evidence. First, similar to the B cell of the adult mammalian SVZ (Merkle et al., 2004; Merkle et al., 2007), boundary astrocytes are directly descended from radial glia the f irst genuine stem cells of the developing CNS (Hartfuss et al., 2001; Levitt and Rakic, 1980; Noctor et al., 2001). Evidence for this descent is clear based on our early morphological analyses of mouse whisker barrel boundary astrocytes. Very early on, o bservable only during the first weeks of postnatal development, these unique cells echo one of the hallmarks of radial glial morphology
32 they secure a direct connection to the pial surface by way of a lengthy radial process. The direct descent from radia l glia is significant because it hints at the possibility that the key set of intrinsic genetic and epigenetic factors required for stemness has been transferred more or less completely to boundary astrocytes (Sakakibara and Okano, 1997). Just as the dire ct descendant of the B cell the C cell is more primitive than its neuroblast progeny (Doetsch et al., 1999), might the boundary astrocyte also maintain a more primitive, stem cell like quality than other more mature astrocytes? At this point, this asp ect of our argument is more of a thought experiment than anything else. Although clearly not definitive, we are intrigued by this genealogical parallel and what it might mean regarding stemness in the population of boundary astrocytes. It would be presu mptive to state that simply having a lineage relationship with radial glia predicates stemness. After all, ventricular ciliated ependymal cells are lineage associated with radial glia (Spassky et al., 2005) and both our own laboratory (Laywell et al., 200 0) and others (Doetsch et al., 1999; Chiasson et al, 1999) have convincingly ruled out the possibility of an ependymal neural stem cell within the brain. Interestingly, Frisen and collaborators have recently presented compelling evidence that ependymal ce lls, in contrast to the situation within the brain, may mediate a reactive cytogenesis following spinal cord injury (Meletis et al., 2008). Nonetheless, focusing specifically on the brain, additional evidence presented by Laywell and colleagues bolsters o ur argument that these unique cortical boundary astrocytes might have the capability of displaying neural stem like character. Laywell and coworkers challenged astrocytes, isolated from various regions within the late embryonic, early postnatal, and adult mammalian CNS, with the neurosphere assay; an in vitro bioassay used to test
33 these tissues for the presence of neural stem like cells (Rietze and Reynolds, 2006). Remarkably, the authors found that until the close of the second postnatal week, the cerebr al cortex harbors a population of astrocytic cells, which display the neural stem cell attribute of neurosphere formation (Laywell et al., 2000). The expression of the germinal ECM undergoes a dramatic shift precisely in the same time period that correlat es with cortical neurosphere formation. Initially expressed broadly at high levels, (Figure 1 5, lower inset) the matrix is progressively spatially restricted until it exclusively defines the adult SVZ. (Figure 1 5) Put another way, cortical boundary as trocytes retract their radial glial like connections to the pia and undergo other aspects of terminal differentiation in precisely the same interval of time that coincides with the end of the critical window for cortical neurosphere formation and the withd rawal of the germinal ECM into the adult germinal center (Cooper and Steindler, 1989). Is the boundary astrocyte the neurosphere forming cell of the neonatal cortex? As we stated above, the boundary astrocyte has never been directly isolated and assessed for neural stemness; however, taken together, these two lines of inquiry strongly suggest that given the proper conditions, boundary astrocytes may represent a unique cortical / CNS potential stem cell. Above and beyond their radial glial ancestry, the idea that cortical boundary astrocytes might be potential stem cells is motivated by the second key feature at the center of this perspective the highly unique makeup of the ECM environment in which neural stem/progenitor cells are found. There is a gro wing body of literature that suggests a connection between the stemness of a cell and the ECM environment in which it is maintained (for review see Alvarez Buylla and Lim, 2004; Nelson and Bissell,
34 2006). For example, consider the adult mammalian SVZ and rostral migratory stream (RMS). The cells of the SVZ and RMS are confined within a uniquely matrix rich environment, strongly expressing the ECM molecules HSPG and CSPG and glycoproteins such as tenascin C (Kerever et al., 2007; Mercier et al., 2002; Thom as et al., 1996). Although the precise role played by environment in SVZ neurogenesis remains unknown, we contend that these surroundings serve to physically sequester the cells of the niche, preserving them in a state of perpetual immaturity. Similarly, boundary astrocytes of the early postnatal cortex themselves reside in an enriched matrix environment, strikingly similar to that of the adult germinal SVZ. Consider for example, early postnatal murine somatosensory cortical boundary astrocytes; these pa rticular boundary astrocytes reside in the spaces, or septae, that will eventually demarcate the walls of the adult murine whisker barrel field. (Figures 1 4A and 1 6) In the neonatal brain, these highly patterned spaces are defined by their intense expr ession of certain lectin binding glycoconjugates and confine the boundary astrocytes, as the neurogenic cells of the SVZ and RMS are confined. If the neural stem cell character of the SVZ B cell is dependent, at least in part, on this unique primordial EC M environment, might such an environment support this same character in the boundary astrocyte? We concede that environment is, as yet, an unproven metric for stemness, especially in these days of the induced pluripotent stem (iPS) cell (Takahashi et al., 2007; Yu et al., 2007) where the intrinsic programming of a cell has cast a long shadow over less well quantified extrinsic influences. Reactive Neurogenic Astrocytes Following Injury or Disease What becomes of the neonatal boundary astrocyte after the down regulation of the ECM boundaries? Might there be a carry over of these interesting cortical astrocytes in
35 the adult brain? One of the main hallmarks of a lesion to the adult CNS is the glial scar (for review see Silver and Miller, 2004). Interestin gly, glial scarring is associated with an enriched ECM, highly reminiscent of the adult SVZ and to neonatal pattern associated boundaries. Similar to boundary astrocytes, astrocytes within and surrounding the lesion, referred to as reactive astrocytes, ar e identified by exaggerated or hypertrophic GFAP expressing processes. Additionally, reactive glia begin to express many of the same markers commonly associated with the neurogenic stem and progenitor cells of the SVZ, and to a lesser extent, the boundary astrocytes of the neonatal cortex. Further, a few only those immediately adjacent to the lesion itself begin to secrete the matrix proteins that define the lesion. At this point, it is crucial to note that this discussion pertains exclusively to a u nique subpopulation of reactive astrocytes. Only those reactive astrocytes that directly circumscribe the lesion and adopt the aforementioned profile cellular immaturity, proliferation, and ECM production are germane. (Figures 1 4C and 1 7) Previous ly, it has been postulated that the re expression of these primitive markers suggests these select reactive glia are potential stem cells (Steindler et al., 1998). Further, we propose that the boundary astrocytes and the specific reactive astrocytes of in terest are similar, lineally related cell types. Both cell types present an immature astroglial phenotype within a common matrix environment, suggesting that both the intrinsic and extrinsic factors required for stemness are accounted for. The Gtz labora tory (Buffo et al., 2008) has recently proven this postulate correct in an elegant article describing the neural stem like attributes of reactive glia both in vitro and in vivo.
36 The notion of astroglial scar formation and the reappearance of boundary mol ecules intimate the possibility that brain injury or disease might re induce normally highly controlled developmental programs for neurogenesis. Penetrating brain injuries result in astroglial scars that exhibit an enhanced expression of ECM proteins incl uding tenascin C and various proteoglycans (Laywell et al., 1992; McKeon et al., 1991; Brodkey et al., 1995). (Figure 1 7A and inset) It is now accepted that within the glial scar, there are cycling cells that exhibit an enhanced expression of astrocytic GFAP (Buffo et al., 2008). These recent results synergize with earlier studies showing reactive astrocytes exhibiting an enhanced expressed of the neural stem cell marker nestin (Lin et al., 1995). There is now no question that following brain injury or disease, the release of growth factors, cytokines, and other factors from at risk or dying cells, as well as vascular and immune related elements, may support a reactive cyto or neurogenesis. In the report by Buffo and colleagues, the authors combined a powerful new conditional GLAST transgenic mouse as well as astrocyte targeted lentiviral vectors. In doing so, they were able to discern cortical reactive astrocyte de regulating rtex. The authors showed that penetrating cortical lesions induced proliferation of the local, resident astrocyte population that contributed to reactive astrogliosis, do not proliferate, and that the adult cerebral cortex mature astrocytes lack expressio n of GFAP, nestin, vimentin, and tenascin C. However, these proteins are contained in some reactive glial cells, some of which proliferate, and in more immature glia such as radial glia and postnatal glial progenitors. Because our fate mapping analysis n ow reveals that reactive astroglia derive from mature astrocytes, these data suggest that astrocytes exposed to injury
37 may indeed resume properties of glia present at earlier developmental stages (Buffo et al., 2008). Whether these cells are in fact linea ge associated with radial glia, boundary astrocytes, or B cell astrocytes remains to be determined. However, their common developmental behaviors, including an ability to maintain proliferation and execute neurogenic programs, along with distinct developm ental molecular expression patterns, suggests that they hold promise for successful CNS regeneration as well as exuberant growth, e.g. neoplasia (elaborated on below). We propose that the boundary astrocyte, the B cell astrocyte, and the reactive astrocyt e have similar if not identical phenotypes in a common matrix environment. Abnormal Astrocytic Neurogenesis Following Oncogenic Transformation Why should a cell that is responsible for protecting and facilitating neural function, e.g. by generating all o f the cells of the brain during development or contributing to adult neurogenesis, change its job and contribute to tissue overgrowth? Stem cells do have an innate purpose of generating tissue, and there is no question that altered oncogene expression, lo ss of tumor suppressor genes, epigenetics and environmental carcinogens can result in such potent cells changing their nature and generating too much tissue. As discussed by Weissman and others (Passegue et al., 2003), it is also possible that cancer init iating cells take on stem cell characteristics rather than starting out as stem cells gone awry, but as discussed above, certain astrocytic cells clearly exhibit a variety of structural and behavioral phenotypic changes during development and injury that c reates the potential to become hyperplastic (Zheng et al., 2002). Clonal neurospheres derived from the human brain exhibit a profound heterogeneity of stem and progenitor cells, when derived from the neurogenic SVZ and hippocampus (Suslov et al., 2002). Along with their ardent responsiveness to changes
38 in growth conditions, this could likewise lead to distinctive examples of cell lineage diversity of their progeny following oncogenic transformation; e.g. the varieties of cell phenotypes seen in the gliom as, especially glioblastoma. (Figures 1 4D and 1 8C, D) Glioma stem like cells were originally discovered in our studies that exploited the same neurosphere culture approaches used for studying normal neural stem cells (Ignatova et al., 2002) and these f indings were corroborated and expanded upon in numerous studies that followed showing the presence of stem cell like, tumor initiating cells in human GBM (Galli et al., 2004; Hemmati et al., 2003; Singh et al., 2004). In our study, as seen in studies of n ormal adult human brain stem/progenitor cells, cells expressed transcripts and proteins associated with stemness including tenascin C, notch and nestin; a transcriptome indicative of common programs for growth and differentiation between normal SVZ B cells reactive astrocytes, and astrocytes found within gliomas. (Figure 1 8A, B) To this end, it is interesting that an anti tenascin C approach has been developed for treating human gliomas (Reardon et al., 2007). In a general sense, the underlying functi on of any stem cell, be it a potential stem cell (Potten and Loeffler, 1990) (e.g. a boundary or reactive astrocyte) or an actual stem cell (e.g. a B cell) is to generate tissue; however, concomitant with this function is an inherent vulnerability. Active cell division opens a cycling cell up to the possibility of accumulating mutations and aneuploid divisions the consensus inroads to oncogenic transformation. With reactive glia, we are presented with a primitive, stem like cell; a cell fully capable of driving cytogenesis (Brodkey et al., 1995) but outside of the self governing confines of the adult neurogenic niches (Petreanu and Alvarez Buylla, 2002).
39 In this final section, we explore the possible consequences of reactive astrocytic cytogenesis speci fically related to brain cancers and the cancer stem cell hypothesis. Embryonal Rest theory of cancer. Based only upon his own detailed histological analyses, Virchow apprecia ted similarities between the cell types present in certain adult tumors and those of the developing embryo. These likenesses lead to the conclusion that cancers were the inevitable byproducts of displaced embryonic tissues. In 1875, his student, Julius C ohnheim, extended and clarified this hypothesis. Cohnheim postulated that adult tissues harbor a residue of dormant embryonic cells, or embryonal rests. These resting, otherwise silent cells, when made active, hold the potential to develop into cancers. These brilliantly intuitive ideas, first articulated a century and a half ago, speak directly to modern discussions of the cancer stem cell hypothesis. There is debate within the cancer field concerning the existence and nature of so called cancer stem cells. (for review see Passegue et al., 2003; Adams and Strasser, 2008; Vescovi et al., 2006) There are those who, similar to Virchow and Cohnheim, argue that cancer is the product of an initially stem like cell gone awry. Although the current literatur nonetheless resident, genetically aberrant tissue specific stem cells. Alternatively, investigators have also entertained the idea of a once mature, newly dedifferentiated cell that has in t he process of aberrant dedifferentiation, acquired stem cell characteristics. However, considering this perspective on stem like reactive glia, a third option emerges: rather than the novel acquisition of stem cell characteristics, in reactive
40 glia we may be dealing more accurately with the reacquisition of dormant stem cell properties. Further, because of this trait, we contend that aberrant reactive glia may serve as a novel biological substrate for brain cancer initiation and provides a yet unexplored therapeutic target for brain cancer research. Bachoo and colleagues (Bachoo et al., 2002) have inadvertently provided strong evidence for a reactive astrocytic tumor initiating cell. Working to elucidate the molecular mechanisms underscoring the most co mmon and most devastating adult brain cancer grade IV astrocytoma, or glioblastoma multiforme (GBM) these investigators chanced upon a fundamental mechanism used by mature astrocytes to preserve terminal differentiation. The authors reasoned, based u pon the genetic profile suppressor and the EGF receptor) might be responsible for oncogenic cellular transformation and resultant gliomagenesis. Remarkably, beginning initia lly in mature astrocytes with a dual knockout, the authors found that the combined loss of p16INK4a and p19ARF resulted in dedifferentiation back to a more primitive, putatively stem like state. When induced by the forced over expression of the EGF recept or, INK4a / ARF / transgenic knockout mice developed a GBM like tumor. Ironically, Bachoo and investigators perhaps wrongfully concluded that the driving force behind the tumor formation was the synergy of the three genetic mutations rather than the a berrant astrocyte cell of origin. As the Gtz laboratory (Buffo et al., 2008) would not confirm reactive astrocytic cytogenesis for another six years we certainly cannot fault these authors such a conclusion, however, in light of current findings, we woul d like to offer a reinterpretation of these data. We challenge that Bachoo and coworkers essentially
41 generated an aberrant reactive astrocyte. They clearly established that p16 and p19 signaling plays a vital role in preserving terminal differentiation i n mature glia. If these loci are compromised, the affected astroglia dedifferentiate back to a more primitive stem like state. In the context of a lesion, the ECM surroundings presumably concentrate growth factors around these de regulated reactive astro cytes. In conjunction, secondary EGFR expression activation exacerbates the proliferative potential of these cells, which in this case, becomes highly problematic. Ultimately, these cells are induced into a highly mitotic state, proliferating in an uncon trolled manner that ultimately manifests as glioma. Our reinterpretation of the conclusions drawn by Bachoo and colleagues, namely that certain reactive astrocytes may serve as a substrate for oncogenic transformation, inflammatory as it may be, is not w ithout precedent in the literature. The first line of support coincidently comes from an independent follow up study to the Bachoo report. Using the same p16 / p19 / EGFR High genetic lesion model introduced in the Bachoo study, Ligon and colleagues demonstrated that ablating the basic helix loop helix transcriptional repressor protein Olig2 in these already genetically aberrant mice, successfully abrogated the onset of gliomagenesis (Ligon et al., 2007). Such a robust result is powerfully striking i n and of itself, however, additional work championed by Chen and co workers the following year brings these results in line with our aforementioned reinterpretation. Chen and colleagues presented a careful examination of the role played by Olig 2 in react ive astrogliosis after cortical injury (Chen et al., 2008). Strikingly, when these investigators ablated Olig 2 specifically from the GFAP+, reactive astrocyte compartment of the lesion, they noted a significant decrease in the
42 proliferation of these lesi on associated reactive glia. Moreover, the authors illustrate that the once proliferative reactive astrocytes affected by the Olig 2 ablation are the same nestin+, vimentin+ potentially stem like reactive glia at the center of the argument presented here. Thus, the same genetic lesion that completely abolishes glioma tumor formation in the p16 / p19 / EGFR High animal, confers a similar suppression of proliferation specifically to reactive astrocytes. Fraser and colleagues provided a second line of s upport for our re analysis of the Bachoo findings (Fraser et al., 2004). The authors presented a careful analysis of another genetic lesion common to many cancers including glioblastoma loss of the tumor suppressor gene PTEN. Remarkably, PTEN ablation specifically from within the GFAP expressing astrocyte compartment generated primitive (nestin+, vimentin+), putatively stem like reactive astrocytes with a significantly enhanced proliferative potential, along with hypertrophy of cells and enlarged brains As noted by the authors, altering genes like PTEN can, constellation of specific gen etic aberrations commonly associated with glioma tumorigenesis, bolstering our reinterpreted hypothesis and presenting them as likely candidates for oncogenic transformation. We have chosen to present the aforementioned study and our reinterpretation in the interests of scholarly provocation. It suggests the possibility that a reactive astrocyte, a cell that has been induced back to a more immature state, could acquire the constellation of mutations necessary for oncogenic transformation. Further, it hi nts at the possibility that such an aberrant reactive astrocyte could be bolstered by extrinsic
43 reactivity and the acquisition of the cancer initiating phenotype is genet ically engineered rather than chanced upon spontaneously. There is an interesting clinical observation linking prior traumatic brain injury to the onset of certain brain cancers later in life. Presumably, a few of the reactive glia associated with such a n injury could provide the biological substrate for transformation. Hasegawa and colleagues have made an instructive attempt to bring clarity and weight to such a clinical correlate of aberrant reactive astrocytic tumor initiation (Hasegawa and Grumet, 20 03). The authors cytogenesis should be the product of an interplay between intrinsic and extrinsic factors. As such, the output of a tumorigenic reactive astrocyte might be bols tered if presented with the germinal like environment of a glial scar and dampened when presented with the unsupportive environment of a healthy CNS. The authors transplanted an astrocytic rodent glioma cell line (C6 glioma) into control and into previous ly contused rodent spinal cords. Invariably, within only 6 weeks of engraftment, the previously injured animals presented with hemorrhagic tumors whereas the uninjured controls did not. The reactive injury associated environment thus provides a molecular niche that appears conducive for tumor growth, again containing cellular and molecular elements in common with the normal tissue generating environment of CNS developmental (boundaries) and adult (SVZ and astroglial scar) neurogenesis. In conclusion, we stem cell literature while attempting to make the case for two interrelated hypothetical ideas. First, stemness is the result of an inextricable link between seed and soil
44 between cell and combination of cell intrinsic developmental programming and the primordial is present within neonatal CN S patterning boundaries, the adult SVZ, and the penumbra of an injury to the CNS only those specific regions to which we attribute the presence of neural stem or stem presented our interpretat ion of a possible consequence of reintroducing a tissue ontogeny for glioblastoma; beginning from a sub population of stem like boundary astrocytes that may subsequen tly re emerge as transformed stem like reactive glia within the adult brain. Ultimately, the re emergence of these immature, aberrant stem like astrocytes may regrettably offer a substrate for neoplastic transformation. Thus, because these three cell typ es: the neonatal boundary astrocyte, the post natal reactive astrocyte, and the astrocytic tumor initiator may actually be three separate manifestations of essentially the same cell, they may represent future targets for therapeutics aimed at encouraging a ppropriate reactive neurogenesis following injury or disease and limiting the extent of neoplastic transformation at the root of gliomagenesis. Glioma Invasion and the Extracellular Matrix Diffuse infiltration of tumor cells throughout the otherwise heal thy brain is a central hallmark of the pathology of high grade astrocytomas (such as glioblastoma multiforme, GBM, WHO grade IV astrocytoma) (Louis et al., 2007). In the most severe cases, microscopic insinuation of the tumor can involve large portions of the brain, spreading throughout the primary tumor bearing and contralateral hemispheres. (DeAngelis, 2001) The products of this extensive invasion are diffuse collections of residual tumor
45 cells that go essentially untreated and remain to perpetuate the malignancy by contributing to treatment resistance and tumor recurrence (Glas et al., 2010). Conversely, certain lower grade gliomas as well as non neural metastases to the brain tend to form focal, space filling, but non invasive lesions. The biologica l underpinnings for these two distinct behaviors invasion versus non invasion remain poorly understood. The list of intrinsic regulators of invasion is extensive and includes factors related to cell cell adhesion, proteolysis, cell substrate attachmen t, and overall cellular motility that all contribute to the innate capacity of a cell to propel itself through tissue. However, the extrinsic elements involved, specifically the interplay between the invading cells and their surrounding microenvironment, are clearly also critical, however far less extensively studied. Such a complex feature of the overall tumor pathology as invasion is arguably, better thought of as a consequence of a mutual exchange between the motile cell populations within a tumor and their immediate microenvironment. Interestingly, the tumor extracellular matrix (ECM) is physically positioned at the junction between the tumor cells themselves and the cells within their surrounding environment, and as such, may be uniquely situated to serve as a critical regulator of these tumor environmental interactions. A central problem that hinders any investigation into invasion is the lack of pathologically accurate animal models of human glioma. The most commonly used models of high grade, hu man astrocytoma rely on the transplantation of human glioma cells, which have been propagated in vitro extensively over the last 4 decades (Pontn and Macintyre, 1968). These classic cell lines have become the industry standard cellular models of human gl ioma both in vitro and in vivo, have been at the center of our
46 efforts to understand the basic biology of glioma invasion (Viapiano et al., 2005; Zhang et al., 1998), and unfortunately sorely underrepresent tumor invasion. Similar to non neural metastases 87MG, U 251MG, and the like) generate aggressively expanding, self contained masses of cells that physically exclude resident neurons and glia, but do not invade into the adjacent tissue. In the absence of robust, pathologically accurate models of the human disease, the biological determinants underlying brain tumor invasion or non invasion remain poorly understood. In response, our laboratory has developed a library of novel human GBM cel l lines, which demonstrate remarkably invasive phenotypes in vivo. When transplanted into the brains of immune compromised mice, the resulting gliomas circumscribe blood vessels, invade within the sub pial space, and demonstrate a strong preference for lo ng distance, single cell migration along myelinated fiber tracts. By invasive glioma models, we have been able to re examine a suite of cellular and molecular mechanisms, s pecifically associated with the tumor extracellular matrix (ECM), and overall tumor microenvironment that appears to have a profound influence over brain tumor invasion. Chondroitin sulfate proteoglycans (CSPGs) are a diverse family of extracellular matr ix (ECM) molecules. Each family member consists of a core protein, for which it is named, covalently linked to at least one long chain chondroitin sulfate glycosaminoglycan (CS GAG) polysaccharide (Galtrey and Fawcett, 2007). CSPGs play critical roles du ring central nervous system (CNS) development as arbiters of neural circuitry pattern formation. They are thought to serve as molecular barriers
47 against the movement of cells across the junction of two adjacent emerging structures. (Cooper and Steindler, 1986; Steindler et al., 1988; Snow et al., 1990) Additionally, CSPGs are a major component of the unique ECM microenvironment of the germinal centers of the embryonic and adult mammalian brain. They contribute to the specialized milieu that fosters the immense proliferation and cytogenesis required of the developing CNS, and maintain this association into adulthood within the rare, persistently cytogenic niches of the adult brain. (Gates et al., 1995; Thomas et al., 1996, Silver and Steindler, 2009) Th ere is an indication that a unique variant of the core protein of one member Brevican of a subfamily of CSPGs referred to as the Lecticans (which also consists of Aggrecan, Versican, and Neurocan) may play a role in brain tumor invasion. (Zhang et al. 1998; Viapiano et al., 2005) However, the CS sugar side chains, which decorate these molecules, and are responsible for their potently inhibitory properties during development, have never been addressed in the context of brain tumor invasion.
48 Figu re 1 1. An astrocytic neural stem cell (NSC) derived neurosphere. A fluorescent micrograph of a once free floating neurosphere in transition to adherent culture conditions demonstrates the multipotential progeny of the sphere forming NSC. Astrocyte and neuronal progeny are illustrated by Glial III Tubulin (Red) immunostaining respectively. Cellular nuclei are visualized by the Hoechst stain (Blue).
49 Figure 1 2. A synopsis of the research that has lead to the current insight into the identity and activity of the astrocytic neural stem cell (NSC). A cartoon depiction of the relationships between the various members of the astrocytic family highlights each of the candidate astroglial cell types that have been considered as the neural stem cell. Also, from a developmental perspective, it is interesting to note that the diverse functionality of the stem like ependymoglial cell of cold blooded vertebrates appears to have been parceled off into individual ast rocytic cells within the mammalian CNS.
50 Figure 1 3. Stem cell characteristics
51 Figure 1 4. Cartoon depicting the four examples of astrocytic stem cells, in vivo and in vitro, that are the focus of this review. (A) Developmental Boundary Astrocy whisker barrel system during early postnatal development. Even though it is predicted that these cells would be both multipotent and clonogenic, and give rise to neurospheres (arrow to the central culture dish), this is as yet uproven. (B) Adult Neurogenic Astrocytes. Multipotent astrocytic stem cells are present in the adult neurogenic niches of the subventricular zone around the lateral ventricle as well as in the hippocampus. These cel ls have been shown in numerous studies to exhibit stem cell characteristics including self renewal, neurosphere generation, and the ability to give rise to all three neural lineages: oligodendrocytes (black cell in center of figure); neurons (pink cell); a nd astrocyes (green cell). (C) Reactive Astrocytes. Injury associated astroglial scar astrocytes attempt reactive neurogenesis following, in this case, a penetrating injury in the adult brain. These cells also exhibit multipotency, giving rise to neurosp heres and all three neural lineages. (D) Tumorigenic Astrocytes. In gliomas, including in this example, a cortical glioblastoma, tumor initiating astrocytic cells not only may give rise to the tumor mass, but also contribute to invasion and metastasis fo llowing their migration to disparate sites. Tumorigenic astrocytes also have been shown in numerous studies to give rise to multipotent neurospheres, thus exhibiting stem cell attributes including the contribution to neural cell lineage diversity as well as self renewal
52 Figure 1 5. Immunofluorescence for tenascin C (main figure) and chondroitin sulfate proteoglycans (lower inset) in the neurogenic subventricular zone (SVZ) surrounding the lateral ventricle as seen in a coronal section through the ad ult mouse forebrain. In contrast to the late embryonic brain (E17) (lower inset) where chondroitin sulfate proteoglycans and other extracellular matrix (ECM) molecules including tenascin C are intensely expressed in both the SVZ and ventricular zone (regi on between star in the lateral ventricle and arrow), as well as in the overlying cortex, there is little labeling in surrounding structures including the striatum, septum, and subcortical white matter. Upper Inset shows a single adult human brain neurosph ere, derived from an SVZ multipotent astrocytic stem cell (MASC or B cell) immunostained for tenascin C. There is a dense tenascin C matrix surrounding cells of this cultured neurosphere. (Figure adapted, with permission, from Gates et al., 1995; Thomas et al., 1996, and Suslov et al., 2002)
53 Figure 1 6. Boundary astrocytes and their expression of developmentally regulated molecules, e.g. tenascin C, cordon off developing structures and functional units through out the neuraxis. In the example pres ented here, somatosensory cortical barrels are demarcated by boundary astrocytes and ECM molecules. (A) In a flattened tangential section through layer IV of the postnatal day 6 mouse somatosensory cortex, immunoperoxidase staining of the ECM glycoprotei n tenascin C reveals boundaries around all five rows of posteromedial and anterolateral subfield whisker barrels Each barrel is 200 (B) GFAP immunolabeling of the postnatal day 6 barrel field reveals astrocytes and their processes that distribute in boundaries around forming barrel un its. A single astrocyte can be seen (arrow) within a boundary (future inter barrel septum) between two barrels. Barrel hollows (two transected blood vessels appear in two hollows) exhibit less GFAP staining, except for sparse transected radial glial proc esses that are more concentrated in the boundaries (appear as immunolabeled punctae). (Adapted, with permission, from Cooper and Steindler, 1986)
54 Figure 1 7 Brain lesions induce certain reactive astrocytes to proliferate and up regulate development ally regulated ECM proteins including tenascin C in response to injury. (A) Immunofluorescence for the tenascin C glycoprotein following a penetrating injury to the adult mouse brain (3 day survival) results in up regulation of this ECM protein just arou nd the injury cavity. Inset in (A) shows in situ hybridization for tenascin C mRNA in the lesioned cortex 3 days following a stab injury, just around the injury cavity. (Adapted with permission from Laywell et al., 1992) (B) Glial fibrillary acidic protein (GFAP) immunoperoxidase and tritiated thymidine autoradiography in the cerebral cortex of a postnatal week 3 mouse following a penetrating stab lesion (lower left side of the figure). In addition to birthday identified vascular elements, GFAP+ rea ctive astrocytes (brown) near the injury site are also found to proliferate in response to the stab wound (black autoradiographic signal over their nuclei), that along with their expression of tenascin C, (A) and (B) suggest possible attempts at reactive n eurogenesis. (Adapted, with permission, from Laywell and Steindler, 1991)
55 Figure 1 8. Human gliomas, both in situ and in vitro, express reactive and neurogenic astrocytic cell markers. (A) and (B) Tenascin C protein immunoperoxidase (A) and mRNA following in situ hybridization (B) show expression of the ECM protein in fixed tissue specimens from a human anaplastic astrocytoma. The dense areas of immunoperoxidase labeling in (A) presumably correspond to clusters of tenascin C riboprobe positve cell s from the same tumor specimen shown in (B). (C) Two attached neurospheres derived from a human glioblastoma specimen indicate the diversity of cell types generated within III tubulin (green) and GFAP (blue)) III tubulin+ cell from one of these spheres exhibits a morphology suggestive of migratory behavior as seen in invasive cells observed from these specimens (unpublished observations, and see migratory cells as depicted crossing the corpus c allosum in Figure 1D). These glioblastoma derived neurospheres also express tenascin C, in addition to other markers associated with normal astrocytic stem cells. (Figure adapted, with permission, from Ignatova et al., 2002), Scale bar in C
56 Table 1 1. Neural Stem Like Astrocytes Cell Type Definition Key References Boundary Astrocyte A cell that expresses GFAP as well as developmentally regulated ECM molecules such as various proteoglycans (CSPG, HSPG) and glycoproteins (TN C). Obs erved throughout the neuraxis at the interface between two functionally different structures during pattern formation. Cooper and Steindler, 1986 Snow et al., 1990 Steindler, 1993 Gonzalez et al., 1993 Miskevich, 1999 SVZ Neurogenic Astrocyt e A cell with morphological and immunophenotypic characteristics of astrocytes. Although a precise surface antigen profile is unknown, these cells are known to express nestin, GFAP, and certain ECM molecules. Reside within the SVZ and contribute to olfac tory neurogenesis in vivo and give rise to clonal, self renewing multipotent neurospheres in vitro Doetsch et al., 1999 Chiasson et al., 1999 Laywell et al., 2000 Garcia et al., 2004 Reactive Astrocyte An astrocyte that up regulates GFAP, nest in, and developmentally regulated ECM proteins in response to injury or disease. Can respond by proliferating and may become neurogenic/cytogenic following exposure to particular environmental conditions both in vivo and in vitro Reier and Houle, 1988 Laywell et al., 1992 Steindler and Laywell, 2003 Silver and Miller, 2004 Buffo et al., 2008 Tumorigenic Astrocyte A transformed astrocytic cell found within human glioma, transgenic mouse models of glioma, or in spontaneously transformed roden t SVZ astrocytes, that exhibits multipotency in vitro and contributes to tumorigenesis in vivo Uhrbom et al., 2002 Ignatova et al., 2002 Bachoo et al., 2002 Hemmati et al., 2003 Galli et al., 2004
57 Table 1 1. Continued Cell Type Definition Key References Qui nones Hinojosa et al., 2007 Siebzehnrubl et al., 2008
58 CHAPTER 2 CHONDROITIN SULFATE PROTEOGLYCANS POTENT LY INHIBIT INVASION AND SERVE AS A CENTRAL O RGANIZER OF THE BRAI N TUMOR MICROENVIRONMENT Here, we present a re examination o f brain tumor invasion, with a specific emphasis on the microenvironment of the tumor, through a comparative study of non invading glioma models with several, newly generated, diffusely infiltrating models, derived from primary human glioblastomas (hGBM L0 hGBM L1, and hGBM L19). We show that the expression of full length, glycosylated CSPGs was strictly associated with non invading tumors, whereas diffuse infiltration only occurred in the relative absence of proteoglycans. Moreover, we demonstrate that the abundantly present, tumor derived CSPGs within the non invading tumor mass induced dramatic cellular alterations and architectural rearrangements within the resident glia (astrocytes and microglia) surrounding and embedded within the lesion. These str omal cell alterations were essentially absent from the CSPG deplete microenvironment of invasive brain tumors and may therefore serve as critical regulators of non invasion. Importantly, we demonstrate that tumor invasion can be modulated by adjusting the amount of CSPGs that are present within the microenvironment of a brain lesion, and speculate on a novel mechanism, mediated through the ECM, and the newly identified CSPG specific LAR phosphatase receptor, by which CSPG expressing tumor cells are bound u p within a CSPG rich ECM and cordoned off by a dense wall of reactive glia. In total, our findings offer a novel clarification of the role played by the ECM in regulating brain tumor invasion and hint at the intriguing possibility that the tumor confining abilities of CSPG stimulated reactive astrocytes might be harnessed in a unique and novel approach toward therapeutic intervention against glioma cell dissemination.
59 Results CSPGs Discriminate Between Invasive and Non Invasive Lesions We compared the tr ansplants of 50,000 cells from 2, non models (U 87MG and U 251MG) with 3 novel human GBM cell lines (hGBM L0, mass of cells with a very clear and well define d edge. These self contained lesions exerted a prominent mass effect, compressing and distorting the surrounding tissue, however, neither line produced any credible invasion beyond the rim of the tumor mass. Conversely, the three novel cell lines (hGBM L 0, hGBM L1, and hGBM L19) each produced diffusely infiltrating tumors. In these cases, there was no discernable edge to the primary tumor mass. Rather, the tumor gradually faded at a microscopic, single cell level into the surrounding tissue. Further, g iven sufficient time (4 6 weeks for hGBM L19, 10 weeks for hGBM L0, and 13 weeks for hGBM L1) each of these cells lines demonstrated long distance invasion, most commonly along myelinated fiber tracts, involving both the initially transplanted and contrala teral hemispheres, up to X mm away from the primary tumor mass. Superimposed upon the distinct cellular pathologies of these two separate brain tumor types, the extracellular matrix (ECM) associated with invasive and non invasive tumors was markedly diffe rent. The non invasive tumors (U 87MG and U 251MG) were each associated with an intense expression of chondroitin sulfate proteoglycans (CSPGs) within the immediate tumor microenvironment. In both cases, CSPGs precisely defined the tumor mass up to and i ncluding the sharp edge of the tumor but no further. Conversely, the invasive tumors (hGBM L0, hGBM L1, and hGBM L19) were all, surprisingly, devoid of CSPGs. (Figure 2 1) Aside from minimal (and inconsistent) expression within the necrotic cores of the primary tumor mass,
60 CSPGs were virtually absent from the microenvironment of the diffusely invasive tumors. Thus, these data attest to the development of a series of novel models of human GBM that each demonstrates a remarkable propensity for long range, diffuse insinuation into the host brain tissue. Further, and to the best of our knowledge, these findings represent the first report of the near complete absence of tumor associated matrix in the context of genuine brain tumor invasion. Specific CSPG C ore Proteins Define the Non Invasive Tumor ECM It is important to consider that, although CSPGs are commonly referred to collectively, as if it were a single molecule, in actuality CSPGs are a diverse family of proteins, related to one another by the lengt hy CS GAG polysaccharide side chains that they all have in common. Further, although both the CS 56 antibody and WFA lectin allow us to visualize the sulfated, CS sugars, neither provides any information on the identity of the core proteins themselves. W e therefore sought to better define the biochemical signature of the non invasive tumor by defining the precise set of CSPG core proteins present within that unique tumor microenvironment. Using immunohistochemistry, we examined a series of 20 m, thin ti ssue sections, bearing the non invading human tumor U 87MG. The non invasive tumors stained robustly for three of four lecticans versican, neurocan, and brevican, as well as the CSPG, phosphacan. Aggrecan, however was markedly absent. In each case, th e individual core proteins were expressed uniformly within the extracellular spaces throughout the entire tumor mass. The matching staining patterns also confirmed that the individual proteins were mixed, homogenously with one another. Further, in parall el with the CS sulfated sugars, the staining for the core proteins abruptly stopped at the tumors outer edge. Interestingly, the absence of aggrecan represented a genuine, and unexpected
61 negative finding, not a failure of the antibody or the immunohistoch emical reaction in general. The peri neuronal nets (PNNs) that condensed around the cortical neurons of the contralateral, non tumor bearing side stained brightly for the missing CSPG, confirming the validity of the immune reaction. Thus, although we wil l continue to refer to the CSPGs collectively, we must keep in mind that the non invasive tumor is actually producing a constellation of overlapping proteoglycans, each with subtly different biochemical properties, which work in concert to define the non i nvasive lesion. Reactive Astrocyte Responses Differentiate Invasive from Non Invasive Lesions Interestingly, invasive and non invasive brain lesions were associated with strikingly different responses from the astrocytes within their immediate vicinities Astrocytes associated with non invading tumors had been forced aside excluded to the edge of the growing tumor. Non invasive tumors therefore, featured a well defined astrocyte circumscription of the tumor mass. Presumably repelled by the growing tu mor, astrocytes had consequently enveloped the mass within a sphere of tightly interwoven astrocytic processes. Conversely, invasive brain tumors left the astrocytic architecture largely intact. These astrocytes formed an elaborate web throughout the ent ire tumor, yet by comparison, remained essentially where they had been prior to tumor implantation. (Figure 2 2) These data suggested that some factor (or factors) associated specifically with non invading tumors and absent from invasive lesions pote ntly repelled the resident astrocyte population to the outskirts of the tumor mass. Because the tumor ECM is so astonishingly different between these separate tumor types, we reasoned that the tumor ECM, specifically tumor associated CSPGs, could be the fa ctor responsible for astrocyte displacement away from a non invading lesion. We therefore tested whether CSPGs, by themselves, could induce astrocyte repulsion. We
62 chose to address this concern in vitro, employing an adaptation of the spot assay first in troduced as a neurite outgrowth assay in the CNS regeneration literature. (Tom et al., 2004a) The spot assay begins with the generation of two opposing gradients of protein within the confines of a single, circular, substrate bound spot. The growth perm issive protein laminin occupies the center of each spot and gradually tapers off approaching the edge. Conversely, the CSPG, aggrecan generates a potently inhibitory rim defining the edge of the spot and tapers off approaching the center. Finally, a uniform coating of laminin is applied to the remaining, unspotted growth surface. For these purposes, we have likened the spot assay to a cross section through the center of a non invading brain lesion. Therefore, the inhibitory proteoglycan rim approxim ates the portion of the CSPG rich tumor mass that makes direct contact with the surrounding astrocytes. We challenged primary, sub cortical mouse astrocytes with spots, fashioned with progressively greater concentrations of CSPGs. Low concentrations of p roteoglycan did not repel the cultured astrocytes; rather astrocytes freely crossed the rim of the spot. However, with increasing levels of CSPGs within the rim, fewer and fewer astrocytes crossed the inhibitory barrier. Effectively, abundant CSPGs induc ed astrocyte circumscription of the spot. It was important that we rule out the possibility that the astrocytes might be repelled not by CSPGs specifically, but merely by the growing abundance of protein at the edge of the spot. However, no astrocyte rep ulsion or circumscription was noted when aggrecan was substituted with equal concentrations of BSA, even at the greatest concentrations of total protein. Thus, these data suggest that CSPGs specifically, are sufficient to induce astrocyte repulsion, and b y extension may explain the astrocyte phenotype surrounding a CSPG enriched, focal tumor.
63 Microglial Responses Differentiate Between Invasive from Non Invasive Lesions We then turned our attention to the microglial component of the tumor microenvironment Unlike astrocytes, which are spatially displaced by CSPG rich, non invasive brain tumors (described above), microglia uniformly populate both types of brain tumors invasive and non invasive. The difference here is in the activation state of the micro glial cells themselves. Microglia associated with an invasive brain tumor are morphologically indistinguishable from their counterparts either in the nave rodent brain or from the microglia that populate the healthy, non tumor bearing regions of the tran splanted host. These microglia, characterized by their ramified appearance, are considered resting, or at their lowest state of activation. Interestingly, this microglial population is clearly interwoven throughout the entire tumor, yet the abundance of surrounding tumor cells is not sufficient to promote any noticeable reaction from the cohabitant microglia. In counterpoint, the situation with non invasive tumors is entirely the opposite. Microglia associated with non invading lesions are all, uniforml y induced to their highest state of activation. Each microglial cell within the confines of the tumor is characterized by the simplified and rounded morphology suggestive of active phagocytosis. Even more importantly, the edge of the non invading tumor a cts as a line of demarcation, separating the ramified microglia outside of the tumor, from the activated microglia within. (Figure 2 3) Thus, microglial activation derives from the non invading tumor itself, rather than the tumor bearing brain tissue in general. We therefore reasoned that some element, specifically associated with non invading tumors and essentially absent from healthy brain tissue or from diffusely infiltrating lesions is at the center of this dramatic microglial phenotypic shift. As noted above, because the tumor ECM is so drastically different between these separate tumor types, we
64 hypothesized that the tumor ECM, specifically tumor associated CSPGs, could be the factor responsible for microglial activation within a non invading l esion. We therefore tested whether CSPGs, by themselves, could induce such robust and consistent microglial activation. We challenged primary, murine microglia with progressively greater concentrations of substrate bound CSPGs deposited evenly across an in vitro growth surface. Low concentrations of proteoglycans maintained the cultured microglia in a state of rest. However, when introduced to increasingly greater levels of substrate bound CSPGs, fewer and fewer ramified microglia could be appreciated. It could be argued that the cultured microglia were activated not by the progressively greater concentrations of proteoglycans specifically, but rather by the growing concentrations of substrate bound protein in general. To control for this confounding f actor, we replaced the CSPGs with escalating concentrations of BSA. The BSA had no discernible activating effects on the microglia, even at the highest concentrations. Thus, because CSPGs alone are sufficient to induce potent microglial activation, their strong presence within non invading lesions, and relative absence from invasive lesions may explain the disparate microglial response between these two distinct tumor types. Reducing CSPG Mediated Inhibition Facilitates Brain Tumor Invasion The near com plete absence of glycosylated CSPGs from infiltrative brain tumors prompted us to question whether the removal (or reduction) of CSPGs from the microenvironment of a non invasive tumor could foster infiltration. Essentially, we sought to test whether the CSPGs themselves inhibit the escape of a potentially infiltrative tumor cell population. We initially addressed this question in vitro by capitalizing on the nature of cultured U 87MG cells to grow in spheroidal aggregates. During the initial growth phas e of the U 87MG culture, the cells grow as an even
65 monolayer. However, above a certain critical density, and typically in the center of the growth surface, the cells coalesce into roughly spherical aggregates. We hypothesized that this tendency for self aggregation in vitro is mediated by the presence of CSPGs and is analogous to the self contained growth of a U 87MG tumor in vivo. Therefore, we cultured U 87MG in the presence of progressively increasing concentrations of the bacterial enzyme chondroitin the long chain carbohydrate moieties from the proteoglycan core protein and has been shown to dramatically reduce the inhibitory properties of these ECM molecules. In a dose dependent manner, as we i fewer aggregates coalesced in the U 87MG cultures. Conversely, U 87MG aggregation was unaffected by the negative control, Penicillinase an alternative bacterial enzyme that has no biological substrate in the U 87MG culture system. (Figure 2 4) Thus, in our preliminary in vitro assessment, CSPGs directly inhibited the dispersal of the otherwise non invasive cell line, U 87MG. These results suggest that CSPGs may actually bind the cells together into a se lf contained mass. Further, this effect is specially mediated by the lengthy carbohydrate side chains that decorate the proteoglycan core proteins. These data indicated the intriguing possibility that CSPGs may actually constrain an otherwise infiltrati ve tumor cell population into a consolidated mass. Therefore, it was critical that we extend this line of questioning into the living animal. Unfortunately, term in vivo experimentatio n. The enzyme is exquisitely temperature sensitive and is rendered inactive in as little as three days at physiological temperatures. (Tester et al., 2007; Lee
66 et al., 2010) We therefore transduced U 87MG cells with lentiviral vectors encoding the genes for chondroitinase ABC (Jin et al., 2011) and/or enhanced green fluorescent protein (eGFP). Initially, we verified the activities of the lenti eGFP expressing cells in vitro in the same aggregation assay as described above. Both tra nsduced cell lines behaved as expected. Because of the continual production of new and active enzyme, cellular aggregation was completely abolished in the lenti ABC cultures. Conversely, the control lenti eGFP expressing cells exhibited a similar growth pattern to either non transduced U 87MG, or the Penicillinase negative control in vivo, two separate cohorts of immune compromised, NOD/SCID animals received unil ateral striatal transplants of either 50,000 U 87MG 87MG eGFP cells (n = 6 each). After three weeks, the tumor bearing brains were excised and examined with immunohistochemistry. The tumors derived from the lenti eGFP expressing cells pre sented as self contained, non infiltrating lesions across the entire rostral caudal axis of the tumor. Strikingly, the lenti exhibit diffuse infiltration at certain positions within the developing tumor. (Figure 2 4) The most robust invasion occurred at the caudal border of the malignancy; however this diffuse pathology did not extend to the main body of the tumor. This incomplete result gave us pause. What could explain this partial shift toward an invasive phenotype? We wondered whether invasion from the caudal extent of the mass was achieved, because intrinsic production of new matrix by the growing tumor. Using the 2 B 6 antibody, which
67 enzyme was active and functional in vivo. There was evidence of chondroitin sulfate most position was an observ able reduction of glycosylated CSPGs detected. Thus, although regionally limited, where the CSPG digestion was most successful, the resulting tumor invasion was considerable. Thus, these findings bolster the initial indications that CSPGs serve, at least in part, to constrain otherwise invasive tumor cell populations, into self contained and focused lesions. Expression of LAR Phosphatase Receptor Distinguishes Non Invasive from Invasive Brain Lesions We thought it interesting that the CSPGs derived from non invasive brain lesions remained so perfectly aligned with the tumor cell mass. More specifically, we never observed the diffusion of these tumor associated proteoglycans away from the outer boundary of a non invasive lesion. This curious behavior st ood in stark contrast with CSPGs that were directly injected into the NOD/SCID mouse brain. Intra striatal injection of high concentrations (700 g/mL) of aggrecan, revealed a relatively dim and widespread cloud of the CSPG. Over a three week period su fficient to develop a sizable non invasive tumor the protein had diffused extensively throughout the brain, favoring white matter pathways, moving across the midline, and deep into the subpallium along the subcortical white matter. Therefore, we wondere d whether non infiltrative tumor cells might bind themselves up within the CSPG rich lesion through an active, receptor mediated coupling. Using immunohistochemistry, we examined infiltrating and non infiltrating tumors for the expression of the leukocyte common antigen related (LAR) phosphatase receptor, a transmembrane receptor that binds CSPGs with high affinity, and contributes to the growth inhibitory properties of CSPGs
68 on axon outgrowth. (Fisher et al., 2011) Interestingly, the LAR phosphatase rec eptor was intensely expressed by non invading lesions, and was entirely absent from diffusely infiltrating tumors. (Figure 2 5) Further, expression of the LAR receptor was perfectly coincident in space with the CSPG expression pattern. Appropriate for a transmembrane protein, the outer membrane of each cell populating the mass could be discerned by its intense staining for the receptor, and the receptor was uniformly expressed throughout the entire non invading tumor. Additionally, in the same way that the edge of the tumor precisely defined the space occupied by the CSPGs, the edge of the tumor also provided the outer limit for the expression of the LAR receptor. Conversely, at immunohistochemical levels, diffusely infiltrating tumors characterized b y the absence of microenvironmental CSPGs were also devoid of LAR receptor protein expression. Thus, these results suggest that that non invasive tumor cells actively bind themselves to their own CSPG rich ECM and in turn, may hold the CSPGs in place. However, these data hold greater implications for the pathogenicity of invasive brain tumors. These data intimate that an inhibitory matrix, providing that one could be established, may not actually be sufficient to constrain a diffusely infiltrative brai n tumor. Addition of CSPGs Helps to Restrict Diffuse Infiltration Since removing CSPGs from a non invasive lesion helped to facilitate invasion, we wondered whether the addition of proteoglycans could be used to constrain an invasive tumor. Although a p otentially appealing option from a translational / therapeutic perspective, our aforementioned LAR phosphatase receptor data immediately brought this possibility into question. Without the appropriate receptor, how could the presence of additional matrix corral an otherwise invasive tumor cell population? We hypothesized that the reactive astrocytes that normally respond to and circumscribe a
69 non invasive lesion might be recruited to constrain an invasive tumor. We initially attempted to establish a reac tive astrocyte barrier around a developing invasive lesion by injecting the tumor cells suspended within a concentrated solution of CSPGs. 50,000 hGBM L19 eGFP cells, suspended in a 700 g/mL aggrecan solution were injected into the striata of four separa te NOD/SCID mice. An additional three animals transplanted with hGBM L19 eGFP cells alone as well as three animals injected with the concentrated aggrecan solution alone were brought to bear as comparative controls. Unfortunately, this paradigm failed to elicit any change in the invasive pathology of the tumor. As described above, rather than remain in position at the engraftment site, the initially highly concentrated aggrecan solution diffused rapidly. This widespread dissemination greatly diminished the potency of the CSPGs at any given point, and as such clearly failed to establish a focus away from which astrocytes might have withdrawn. We therefore reasoned that such a drastic microenvironmental re organization required a tightly focused point of intense CSPG concentration, a highly reactive point of induction, away from which astrocytes would retreat. Coincidently, because the effect we were looking for was the natural consequence of a non invasive lesion on the brain, we reasoned that co transp lanting the infiltrative tumor line, hGBM L19 along with the non invasive tumor, U 87MG, might provide the right amount of matrix, in the right location, and help to clarify the role played by the enveloping reactive astrocyte population. It was important that we control for the possibility that these two tumor lines might proliferate at different rates within the living brain. This was not the case in vitro; however cell division rates in culture may or may not predict proliferation rates in vivo. Obser ving an invasive tumor cell population
70 strictly within the bounds of a larger, non invasive mass might indicate that the non invasive tumor is capable of sequestering the invasive cells. However, we would expect this same result if the growth rate of the infiltrative tumor was greatly outstripped by that of the non invasive mass. Therefore, three separate cohorts of mice (n = 3 each) received intra striatal transplants of 50,000 total tumor cells comprised of the invasive line hGBM L19, transduced with a lentiviral vector encoding firefly luciferase (hGBM L19 f Luc), in addition to the non invasive tumor line U 87MG, expressing eGFP (U 87MG eGFP). One cohort received equal parts invasive to non invasive cells (1:1), the second cohort received double the n umber of invasive to non invasive cells (2:1), and the third cohort received half the number of invasive to non invasive cells (1:2). Additionally, we allowed enough time (3 weeks post transplant) for the tumors to develop, such that an unencumbered invas ive tumor would be capable of long distance infiltration into the corpus callosum, approximately to the midline of the brain. Remarkably, in each case, the non invading mass completely sequestered the invasive tumor cell population. Visualized using immu nohistochemistry, a large, well defined mass occupied the majority of the striatum. The f Luc expressing, invasive cells presented as collective tendrils of cells, swirling within the larger space filling mass. These tendrils approached and, in some case s skirted the edge of the larger mass, however, we did not observe any cells actually escape the confines of the non invasive lesion. (Figure 2 6) Because invasive brain tumors lack the appropriate receptor to bind CSPGs, these data suggests that the env eloping reactive astrocytes may not be mere bystanders in this situation. Rather, these observations imply that the enveloping
71 astrocytic barrier around the outer, non invading lesion may actually serve as a physical barrier against the release of invasiv e cells beyond the edge of the tumor. MMP Profile of Invasive and Non Invasive Lesions We are accustomed to thinking about tumor invasion from the perspective of epithelial cancers, which invade (and may subsequently metastasize) only after the breakdown of an initially present, inhibitory ECM. The idea that this barrier may be essentially absent from high grade glioma represents a significant departure from the generally held view of tumor invasion. The matrix metalloproteinase (MMP) family of proteins specifically MMPs 2, 9, and 14, are considered the agents responsible for the breakdown of this matrix barrier in the context of non neural tumor invasion. (Wolf et rich experimen tal models, this same set of MMPs has been associated with the diffuse infiltration of high grade gliomas. We wondered how these findings might translate to our diffusely infiltrative brain tumors, where the inhibitory matrix is essentially absent. Using gelatin zymography, we examined the soluble protein fraction of tumor bearing hemispheres from animals previously transplanted with either invasive (hGBM L19) or non invasive (U 87MG) tumor cells for functional MMP 2 and 9. In order to minimize the amoun t of potentially confounding rodent host protein present within the protein lysates, we harvested tumors strictly from long term surviving animals, in which the tumors occupied large portions of the excised hemispheres. We also included protein isolated f rom the same hemisphere of a nave, age matched animal as an additional negative control. Interestingly, we found no detectable MMP 2 or 9 activity from protein isolated from either the diffusely infiltrating hGBM L19 tumor, or the nave host brain; howev er, functional MMP 2 was evident within the CSPG enriched, non invasive U 87MG tumor.
72 We confirmed these findings using immunohistochemistry. Representative tumor bearing sections from invasive and non invasive tumor transplants corroborated the strong p resence of MMPs within the non invasive U 87MG tumor as well as their absence from each of our diffusely infiltrating gliomas. These data suggest that unlike epithelial tumors and non invasive glial tumors, diffusely infiltrative gliomas, which are intrin sically free of an inhibitory ECM, do not require MMP mediated proteolysis for successful insinuation into the brain parenchyma. Additionally, these data rule out the possibility that CSPGs may have been produced by the tumor only to be immediately proteo lyzed and presumably, therefore undetected by overly abundant and highly active MMPs. CSPGs Dicriminate Between Inva s ive a nd Non Invasive Human Tumors Because these data support a somewhat contrary model of glioma invasion one without the prerequis ite breakdown of a proteoglycan rich ECM it was critical that we validate our experimental findings with genuine human, neuro pathological specimens. We hypothesized that, similar to U 87MG, low grade (WHO grades I or II), relatively non invasive glioma s would present with a CSPG rich tumor ECM and well defined edges circumscribed by reactive astrocytes. Further, analogous to our experimental hGBM lines, infiltrative gliomas would present with negligible micro environmental CSPGs and widely distributed reactive astrocytes throughout the tumor bearing tissue. We examined thin sections of archival, paraffin embedded samples of tumor bearing tissue from neurosurgical resections using standard histology and immunohistochemistry. In parallel with our experi mental findings, the low grade benign glioma, pleomorphic xanthoastrocytoma (PXA, WHO grade II astrocytoma, n = 5) presented with clean borders and abundant reactivity for the CSPG specific antibody
73 CS 56. CSPG immune reactivity was distributed uniformly throughout the tumor mass, precisely defining the boundary of the PXA, and by default, established a line of demarcation separating the benign tumor from the healthy adjacent brain tissue. (Figure 2 7) We also observed a curious difference in the CSPG ex pression patterns between the human specimens and the experimental U 87MG xenografts. Within a given section, CSPGs were evident within the extracellular space surrounding each cell, uniformly spanning the entire cross sectional area of the tumor. Within the PXAs however, the protein was expressed in ribbons, woven intricately throughout the tumor although not surrounding each and every hyperplastic cell. Additionally, in accord with our experimental results, CSPG expression in high grade, aggressively i nfiltrative GBM (n = 5) samples was negligible. We did observe some subtle CS 56 immune reactivity within pseudo palisading necrotic sites; however, this staining was likely non specific. Regardless, the intensity of the stain in the non invasive PXAs fa r outstripped any questionable staining in the GBM samples. In total, these data help to validate our experimental findings and assert that our novel human glioma cell lines reflect the dominate the literature. Additionally, these data provide critical support for the novel concept that genuine glioma invasion occurs in the absence of a proteoglycan rich inhibitory ECM. Methods Cell Culture Primary human cancer cell cu lture hGBM L0 (derived from a 43 year old male) and hGBM L1 (derived from a 45 year old female), both classified as WHO grade IV astrocytomas, were initially established as
74 described by Deleyrolle and colleagues. (Deleyrolle et al., 2011) hGBM L19, a p ediatric glioblastoma cell line, was derived from a 9 year old male, characterized as a WHO grade IV astrocytoma. The complete account of the isolation and cultivation of hGBM L19 from the original patient tumor specimen is beyond the scope of this report but will be elaborated in a forthcoming article by Scheffler and colleagues. Briefly, the tumor specimen, collected in accordance with the University of Florida Institutional Review Board (IRB), was subdivided into multiple smaller micro explants roughl y 1 mm3. These micro explants were then evenly distributed across a poly L ornithine (15 g/mL, Sigma Aldrich, St. Louis, MO) coated culture surface and sustained by a minimal volume sufficient to suffuse the substrate without enabling the micro explant s to float of N2 growth medium supplemented with fetal bovine serum (FBS, 5% v/v, Atlanta Biologicals, Lawrenceville, GA), the mitogens recombinant human epidermal growth factor (rhEGF, 10 ng/mL, R&D Systems, Minneapolis, MN), recombinant human fibroblas t growth factor (rhFGF basic, 10 ng/mL, R&D Systems), recombinant human leukemia inhibitory factor (rhLIF, 1 L/mL, Millipore, Billerica, MA), and the glycoprotein natural mouse laminin (1L/mL, Invitrogen, Carlsbad, CA). The N2 growth medium itself consi 12 Nutrient Mixture (DMEM/F12, Invitrogen) bolstered by recombinant human Insulin (5 mg/L, Millipore), recombinant human Transferrin (100 mg/L, Millipore), Sodium Selenite (NaSel, 30 nM, Sigma Aldrich), Proges terone (20 nM, Sigma Aldrich), Putrescine (100 M, Sigma Aldrich), Bovine Pituitary Extract (BPE, 2.5 mL/L, Invitrogen), and antibiotic antimycotic solution (abx, 1X, Invitrogen). Taking great care to preserve the attachment of the explanted tissue to the growth surface, the volume of medium was gradually increased
75 over an approximately weeklong period. During this time, tumor derived cells began to emigrate from the micro explants onto adjacent portions of the growth surface and proliferate. At weeks en d, the tissue fragments were discarded and the growth medium was exchanged. Complete medium, including mitogens and laminin was exchanged every other day until the culture reached 85 90% confluence. The cells were then culled and cryopreserved, without further passaging, for future use in subsequent studies. During the course of these studies, the three primary human cancer cell lines hGBM L0, L1, and L19 were each propagated under the following adherent culture conditions. A single cell suspension of 50,000 cells/mL was seeded evenly across a poly L ornithine coated culture surface and nourished by N2 growth medium supplemented by FBS (5% v/v) and the mitogens rhEGF, rhFGF, and rhLIF. Note: the soluble laminin included as a supplement to N2 medium during the initial isolation of hGBM L19 was omitted during these investigations. Medium and supplements were exchanged every other day until the cultures reached 85 90% confluence. Approximately once every 7 days, cultures were passaged using trypsin EDTA (0.25% trypsin, 1 mM EDTA, Atlanta Biologicals) and filtered through a 70 m Nylon cell strainer (BD Biosciences, Bedford, MA) in order to re establish a single cell suspension before re seeding at 50,000 cells/mL in N2 growth medium. Classic human cancer cell culture During the course of these studies, both U 87MG and U 251MG were propagated in Minimum Essential Eagle Medium (EMEM, ATCC) supplemented with FCS (10% v/v ) and abx according to protocols established by ATCC with the following modific ation. Between each passage, the cellular suspension was filtered through a 70 m Nylon cell
76 strainer in order to re establish a single cell suspension before re seeding at 50,000 cells/mL in complete growth medium. Primary murine astrocyte cell culture The method employed herein for the isolation and cultivation of primary astrocytes was adapted from our previous efforts generating astrocyte monolayer cultures from the early postnatal rodent brain (Laywell et al., 2000). Briefly, a block of tissue c ontaining the left ventricles, septum, and bilateral striata was prepared from neonatal (postnatal days P4 to P8) C57BL/6J mice (The Jackson Laboratory, Bar Harbor, ME). Typically, tissues from three to four animals were culled and processed collectively for a single astrocyte monolayer culture. A single cell suspension was prepared from the collected blocks of tissue by way of a series of mechanical and chemical dissociation techniques. The tissue, minced roughly with a sterile razor blade, was incubate d in trypsin EDTA in a 37C water bath for 5 minutes. The cell suspension was then thoroughly triturated through a series of progressively smaller diameter pipettes and filtered through a 70 m Nylon cell strainer. Finally, cells were seeded evenly acros s a 25 cm2 growth surface and nourished by N2 medium containing FBS (5% v/v) and the mitogens rhEGF and rhFGF. Medium and supplements were exchanged every other day until the cultures reached 85 90% confluence. Approximately once every 7 days, cultures were passaged using trypsin EDTA and filtered through a 70 m Nylon cell strainer in order to re establish a single cell suspension before re seeding at 50,000 cells/mL in complete growth medium. Astrocytes were maintained in culture strictly through fou r subsequent passages from their initial isolation. Thus, the astroglial spot assays (see below) were all conducted using cells from these first four initial passages in vitro.
77 Primary murine microglial cell culture The blocks of tissue used to genera te the astrocyte monolayer cultures (described above) also contain a limited population of primary microglia. In order to expand and gain access to this initially minor cellular fraction, we apply a protocol, devised by Marshall and colleagues (Marshall e t al., 2008), that employs the cytokine, Granulocyte Macrophage Colony Stimulating Factor (GM CSF) to selectively drive proliferation within the microglial cell pool. Briefly, either primary or first passage astrocyte cultures were grown to 95 100% conf luence. At this point, the standard N2 growth medium was replaced with microglial proliferation medium (MPM). MPM consists of N2 growth medium supplemented by FCS (10% v/v), abx solution, and GM CSF (20 ng/mL, Stem Cell Technologies, Vancouver, BC). The MPM was exchanged every other day for an approximately weeklong period during which time, phase bright, spherical microglia gradually accumulated either directly atop the astrocyte monolayer or floating within the culture medium. At weeks end, the culture flask was gently agitated at room temperature for 10 15 minutes to loose the cells that were directly attached to the astrocyte monolayer. Lastly, the detached microglia, suspended within the culture medium were culled and processed for further experim entation and analysis. Genetic Modification of Cancer Cell Lines Subsets of U 87MG were transduced with lentiviral vectors encoding the genes for either chondroitinase ABC (LV eGFP). (Generous gifts from Dr. George Smith, Univerisity of Kentucky and Dr. Lung Ji Chang, University of Florida, respectively). 50,000 cells/mL were plated evenly across 6 wells of a sterile 6 well plate. Overnight incubation ensured the firm attachment of the cells to the culture surface while maintaining an appropriately low density (30 50%
78 confluence) for optimal transduction efficiency. The following day, the medium was exchanged with a minimal volume (500 L/well) of virus containing medium, which consisted of comple te growth medium charged with 10g/mL polybrene (Sigma) and the virus solution. Each well of the six was treated separately and two were reserved as experimental controls. The first control a non transduced control was treated to standard growth cond itions, whereas the other was used as a polybrene control. The other five received a progressively increasing viral titer, from 10 to 100 multiplicity of infection (MOI). The plate was removed from the incubator once every hour, for the following three h ours for gentle agitation, then left alone overnight. The following day, additional growth medium was added without removing the initial virus containing medium. The medium was finally exchanged with standard U 87MG growth medium on the third day and eve ry other day thereafter until the culture reached 85 90% confluence. The single well of transduced cells with the highest MOI that demonstrated a growth rate approximately equivalent to the non transduced control was carried forward and expanded for sub sequent experimentation. Aggrecan Laminin Spot Gradient Assay Tom and colleagues originally conceived the spot assay (Tom et al., 2004a) as an in vitro model of glial scarring and neuronal regeneration. In order to apply this assay to the astroglial c ompartment of the non invasive brain tumor microenvironment, we have had to modify both the formulation of the spot solutions as well as the schema for interpreting and quantifying the results. First, regarding the basic assembly and structure of the assa y, we began by treating poly L ornithine (15 g/mL, Sigma) coated, 18 mm glass coverslips with a volatile solution of nitrocellulose (1.6 cm2 fragment dissolved in methanol; Bio Rad, Hercules, CA). As the solvent evaporates away, a
79 uniform coating of nitr ocellulose is deposited across the culture surface, which in turn facilitates the binding of the aggrecan and laminin proteins into a tight circular spot. It is important to note that spots can be prepared without substrate bound nitrocellulose; however, in these cases, the spots tend to dry with irregular, undulating edges, which can confound interpretation and downstream quantification. The opposing gradient spots themselves were established by placing 4, 3 L droplets of an aggrecan (10 250 g/mL, Si gma) and laminin (10 g/mL, Invitrogen) containing solution in DMEM/F12 onto the pre treated glass substrate in a roughly square shaped pattern and allowed to dry completely. Each aggrecan / laminin solution was used to prepare 12 redundant spots across t hree separate cover glasses. Additionally, negative control spots consisting of BSA (25 250 g/mL, Sigma) and laminin (10 g/mL) as well as laminin alone (10 g/mL) were also included and prepared in parallel with the experimental groups. Once the spot s had dried, the entire culture surface was suffused with a solution of laminin (10 g/mL) in DMEM/F12 and incubated at 37C for 3 hours. Finally, the laminin bath was gently aspirated and 50,000 primary murine sub cortical astrocytes suspended in standar d astrocyte growth medium (described above), were distributed evenly across the spot treated surface. The medium was exchanged every other day until the culture reached 85 90% confluence within the non spotted portion of the cover glass. The completed assay was then fixed in warm 4% formaldehyde for 30 minutes and prepared for subsequent immunocytochemical staining. The assay was bathed in blocking solution consisting of fetal bovine serum (FBS; 10% v/v), goat serum (GS, 5% v/v, Vendor Information), ho rse serum (HS, 5% v/v, Vendor Information), and Triton X 100 (0.1% v/v, Vendor Information) dissolved in phosphate buffered saline
80 (PBS, Vendor Information) during an initial 4C, overnight incubation. The assay was then immuno stained at 4C, overnight f or the antigens glial fibrillary acidic protein (GFAP; 1/800, Dako, Glostrup, Denmark) and CS 56 (1/200, Sigma). The primary antibodies were coupled to appropriate Alexa Fluor 488 (1/500, Invitrogen) or 555 (1/500, Invitrogen) conjugated secondary antib odies during a final overnight incubation. Lastly, the assay was overspread with a blocking solution containing the nuclear dye 4',6 diamidino 2 phenylindole (DAPI; 0.1g/mL, Sigma) for 10 minutes at room temperature. The cover glasses were mounted to gl ass microscope slides within a fluorescence protecting mounting medium (Vectashield, Vector Laboratories, Burlingame, CA) and examined using a Leica DMLB standard fluorescent microscope. In order to quantify the spot assay, 6 images were captured around t he CSPG rich rim positions. The total number of astrocytes with DAPI+ nuclei in the outer spot region, which extended processes across the rim and into the penumbra of t he spot were tallied using the Cell Counter plugin for the open source software package ImageJ (National Institutes of Health, USA). It is important to note that, even at the highest concentrations of proteoglycan, a subset of astrocytes consistently grew within the core of the spot where the effective CSPG concentration is far lower than at the rim. As we never encountered this situation in vivo astrocytes growing within the confines of a non invasive tumor these inner spot astrocytes were disregarde d from quantification. Microglial Activation Assay We began assembly of the microglial activation assay by treating 18 mm glass coverslips with poly L ornithine (15 g/mL, Sigma) at 37C for no less than 18 hours. Once this initial coating was establish ed, an additional coating which served as the
81 direct substrate for the assay of laminin (10 g/mL) and the CSPG aggrecan (25 250 g/mL in DMEM/F12) was added over a second 3 hour, 37C incubation. Additionally, negative control conditions consisting of BSA (25 250 g/mL, Sigma) and laminin (10 g/mL) as well as laminin alone (10 g/mL) were included and prepared in parallel with the experimental groups. Three redundant cover glasses were prepared for each experimental and control condition. Once the topcoat was secure, these solutions were gently aspirated and replaced by 50,000 primary murine microglia per well suspended in microglial proliferation medium (detailed above). Media was exchanged every other day for 5 days before a 10 minute fixatio n in warm 4% formaldehyde. The completed assay was blocked in detergent free blocking solution (10% FCS, 5%GS, 5%HS in PBS) overnight at 4C before a 48 hour, 4C immuno staining for the microglial surface antigen CD11b (1/200, BD Pharmingen, San Diego, C A). The primary antibody was coupled to an appropriate Alexa Fluor 488 (1/500, Invitrogen) conjugated secondary antibody during a final overnight incubation and finished with DAPI. In order to quantify the microglial activation assay, an observer, bli nded to the substrate conditions of the assay, first surveyed the entire culture surface of each cover glass with a series of 10 non overlapping images, at 100X magnification. Employing the Cell Counter plugin for the open source software package ImageJ ( National Institutes of Health, USA), the blinded observer was then instructed to score each cell as either ramified or activated based on its unique cellular morphology. For these purposes, ture with three or
82 been essentially spherical, simple cells without elaborate processes. However, in any given visual field, a subset of microglia presented a morphology that was difficult to classify. These so morphological spectrum between the extremes of microglial activation and rest. Because the incidence of incomplete microglial activation in vivo was negl igible, these of the scoring system. In Vitro Tumor Dispersal Assay 50,000 cells/mL of U 87MG were plated uniformly across the wells of an untreated, sterile 6 well plate in standard growth media. Two redundant wells were seeded for each experimental and control condition. Immediately thereafter, either chondroitinase ABC (0.025 U/mL 0.1U/mL) or the negative control enzyme penicillinase (0.025 U/mL 0.1U/mL) was added to the culture media. Complete growth media, including the proper enzyme was exchanged every other day in order to treated wells rea ched 85 90% confluence. Using a Leica DM IRB inverted microscope, outfit with a Leica DFC 300F digital camera, a series of two to four low magnification (25X), phase contrast images of the aggregate occupied area of the growth surface of each well were captured. The Photomerge automation for Photoshop CS4 (Adobe Systems, Inc., San Jose, CA) was then used to reconstruct the entire aggregate occupied area into a single image. Finally, the total number of visible, phase dark aggregates was quantified usin g the Cell Counter plugin for the Image J open source software package (National Institutes of Health, USA).
83 Intracranial Transplantation Surgical procedures All animal procedures were conducted in accordance with protocols previously vetted and approved by the University of Florida Institution Animal Care and Use Committee (IACUC). Adult female NOD/SCID mice (Charles River Laboratories, Wilmington, MA) were induced into a surgical plane of anesthesia using inhaled USP grade isoflurane (2 2.5%; Halocar bon, North Augusta, SC) in oxygen (2L/min). Once the animal was transferred to the stereotaxic apparatus, a sterile field was established and the skull was demonstrated. A burr hole was drilled 0.5 mm rostral and 1.8 mm lateral of bregma. A 33 guage, st ainless steel needle (Hamilton, Reno, NV) was lowered 2.5 mm beneath the surface of the brain and 50,000 cells suspended in 1 L of sterile culture medium were slowly injected over the course of approximately 5 minutes. The needle was held in place for an additional 5 minutes before gentle and measured retraction from the brain. Finally, the scalp incision was closed with sterile staples. Post operative survival time varied for each tumor model under scrutiny. However, in each case, it was important tha t animals were culled before the tumor mass had become so large that the tumor microenvironment could no longer be appreciated. As 87MG and U 251MG survived in a University of Florida vivar ium for no longer than 5 weeks, whereas the primary glioma cell cultures hGBM L19, hGBM L0, and hGBM L1 survived for 4 6, 12, and 13 weeks respectively. Tissue preparation and immunohistochemistry At the close of the post operative period, all animals w ere anesthetized and transcardially perfused with cold 4% formaldehyde (Sigma) in PBS. The intact brain
84 was carefully removed and immediately transferred to a fresh volume of 4% formaldehyde for overnight post fixation. The tissue was then immersed in a solution of 30% sucrose (Sigma) in PBS until equilibrium was achieved and the tissue had sunk to the bottom of its container. A second, overnight immersion in a 1:1 solution of 30% sucrose : O.C.T. Compound (Tissue Tek, Torrance, CA) completed cryo protec tion and readied the tissue for frozen sectioning. Finally, the tissue was embedded in O.C.T. Compound, frozen, and 20 m coronal sections were prepared on a Leica CM 1850 or a Microm HM 505E cryostat. Subsets of these tissue sections were culled for imm unohistochemical analysis against a battery of antigens specific to the experiment and question under study. Tissue sections were immuno stained in detergent containing blocking solution according to the procedure described above for the aggrecan lamini n spot gradient assay. A detailed list of the antibodies employed can be found in the supplementary data. In all cases, primary antibodies were coupled to appropriate Alexa Fluor 488 (1/500, Invitrogen) or 555 (1/500, Invitrogen) conjugated secondary a ntibodies. Quantification of in vivo tumor invasion 20 m thin sections were initially immuno stained according to the protocol described above. The xenografted human glioma cells were visualized in green, using an antibody raised against Human specific Nestin (hNestin, 1/1000, Millipore, Billerica, MA). Additionally, in order to examine the tumor in context, the murine host tissue was III Tubulin (1/1000, Promega, Madison, WI) and nuclei were visualized with DAPI (0.1g/mL, Sigma). The stained sections were mounted onto Colorfrost Plus micros cope slides, (Fisher Scientific, Waltham, MA) overspread with a fluorescence protecting mounting medium, (Vectashield, Vector
85 Laboratories, Burlingame, CA) cover slipped, and examined using a Leica DMLB standard fluorescent microscope equipped with a Spot RT3 CCD camera (Diagnostic Instruments, Inc., Sterling Heights, MI). Using Spot Advanced software, (Diagnostic Instruments, Inc.) a series of three color images were captured at 50X magnification covering at least the tumor bearing hemisphere, if not the entire tissue section. The Photomerge automation for Photoshop CS4 (Adobe Systems, Inc., San Jose, CA) was then used to reconstruct the entire series into a single image. Photomerged images were then imported in Image J and separated by channel. The gre en channel, representing the hNestin+ glioma cells was then inverted generating a black on white image and the threshold was adjusted in order to distinguish the tumor from any non specific background staining. Finally, the Analyze Particles function was used to quantify the total number of discrete particles contained within the tumor. We found that diffusely invasive tumors were made up of several thousand times (~ 2000 7300) more discrete, non contiguous particles than non invasive lesions. Gel atin Substrate Zymography Zymography was performed on the soluble protein fraction isolated from in vivo tumor bearing tissue from animals previously transplanted with either invasive (hGBM L19, X weeks survival) or non invasive (U 87MG, X weeks survival) tumor cells. Novex Zymogram gels were used with the XCell SureLock Mini Cell electrophoresis system as instructed by the manufacturer (Invitrogen). The gel was stained for 4 hours at RT with an aqueous solution of 0.1% Coomassie Brilliant Blue R250 (Serv a Electrophoresis, Heidelberg, Germany) and 16% glacial acetic acid and de stained for 10 minutes at RT in an aqueous solution of 30% EtOH and 10% glacial acetic acid. Areas of protease activity, which appear as clear bands against the dark blue backgroun d of the
86 coomassie blue stained gelatin, were imaged on a FluorChemQ Multi Image III system (Cell Biosciences, Santa Clara, CA) equipped with AlphaInnotech software version 22.214.171.124. The identity of the MMP was determined based on co migration with purifie d recombinant human MMP positive controls as well as molecular weight standards. Human Tissue Collection and Processing All human tissue was acquired and processed in accordance with protocols previously vetted and approved by the University of Florida G ainesville, Health Science Center Institutional Review Board (UF IRB 01). Briefly, 5 m thin sections were prepared from de identified blocks of paraffin embedded human tumor specimens, which had been previously collected from the UF, Shands Department of Neuropathology tissue archives. After air drying overnight at room temperature, the slides were sequentially de paraffinized, rehydrated, and blocked for endogenous peroxidase activity. Optimal staining required 25 minutes of heat antigen retrieval in 1 0 mM Citrate buffer (pH 6.0). The tissue was then immuno stained at 4C, overnight for the antigen CS 56 (1/200, Sigma). Slides were stained using the ABC Elite kit (Vector biotin blocking steps included. Positive staining was detected with DAB (Vector Labs) as the chromogen and hematoxylin 560 (SurgiPath, Richmond, IL) as the nuclear counterstain. Statistical Analysis Data were analyzed using the GraphPad Prizm 5.0 (Graph Pad Software, La Jolla, CA) software package. In all cases, a P value of <0.05 was deemed significant.
87 Figure 2 1. CSPGs discriminate between diffusely invasive and non invasive brain lesions. The development of a series of novel human GBM cell li nes revealed a startling disparity between the matrix content of invasive and non invasive brain lesions. Representative, low ( A ) and high ( B ) magnification fluorescence micrographs of diffusely infiltrative h GBM xenografts reveals a paucity of the EC M protein CSPG. In contrast, ( C D ) CSPGs are present in great abundance in non invasive, strictly expansively growing lesions. ( E ) Quantification of the invasive character of a series of four human GBM cell lines illustrates the tremendous dispariti es between invasive and non invasive tumor phenotypes. (mean SEM, one way ANOVA, Bonferroni Dunn post hoc test, p < 0.0001, n = 6 / group) ( F ) Quantification of the fluorescence intensity of the CSPG specific, WFA Lectin reveals the stark dissimilarit y in the proteoglycan content between invasive and non invasive GBM transplants. (a.u. = arbitrary units, mean SEM, unpaired t test, p < 0.0001, n = 18 / group) Scale bar = 50 m.
88 Figure 2 2. Reactive astrocytes withdraw from the space occupied by CSPG rich; non invasive lesions but persist throughout regions occupied by invasive tumor micrographs illustrate the astrocytic circumscription of non invasive, CSPG heavy lesio ns. Note the scant GFAP+ processes remaining within the penumbra of the non invasive mass. In contrast, representative low (C) and occupation of infiltrative lesions by reactive ast rocytes. CSPG expression level may help to clarify this disparate astrocyte response. (E) Spot gradient assays illustrate the strong aversion that reactive astrocytes demonstrate toward increasing concentrations (F, G) of substrate bound CSPGs. (H) Quan tification of the spot gradient assay emphasizes the inhibitory effects of high levels of CSPGs on reactive astrocytes. (Mean SEM, one way g/mL CSPG groups, n = 72 visual fie lds for 5 100 g/mL CSPG groups) Scale bar: A G = 10 m.
89 Figure 2 3. Microglial activation state differs markedly between invasive and non invasive brain lesions. Representative low magnification fluorescence micrographs demon strate that microglia uniformly populate both non invasive confines of a CSPG rich non invasive lesion demonstrate the simple, spheroidal morphology associated with activation. In co ramified morphology of invasive tumor associated microglia suggests a resting phenotype. (G, H) Microglial activation assays support a direct causal relationship between the microglial activation state and CSPG concentration. (I) Quan tification of the microglial activation assay demonstrates the direct link between CSPG concentration and activated microglial as well as the concomitant indirect link between CSPG concentration and ramified microglia. (Mean SEM, two way ANOVA, Bonferro ni Dunn post hoc test, *p < 0.0001, #p = < 0.0001, n = 30/group) Scale bar: A, D = 50 m; B, C, E H = 10 m
90 Figure 2 4. Reducing CSPG mediated inhibition facilitates brain tumor invasion. (A, B) Tumor dispersal assays demonstrated that U 87MG aggr egation (A) is mediated through the potently inhibitory CS side chains that decorate the proteoglycan core proteins. (C) In a dose dependent manner, addition of chondroitinase ABC, but not the negative control enzyme penicillinase, to the culture medium p revented aggregation and fostered the maintenance of an even monolayer (B) of cells. (Mean SEM, one way ANOVA, Newman transduced negative control cells, (E) lentiviral transduction of U 87MG cells with chondroitinase ABC enables the diffuse infiltration of once non invasive tumor cells. (F) Immunostaining for 2B6 antibody confirmed that U 87MG infiltration occurred within zones of digested CSPGs. (G) Quantification of chondroitinas e ABC mediated U 87MG invasion clarifies the inhibitory effects of CSPGs on tumor invasion. (a.u. = arbitrary units, mean SEM, unpaired t test, *p < 0.05, n = 7 for LV eGFP; n = 15 for LV m
91 Figure 2 5. The CSPG receptor LAR differentiates invasive from non invasive lesions. Representative fluorescence micrographs of (A) non invasive and (C) invasive lesions highlight the disparity in the expression of the CSPG specific LAR phosphatase receptor. (B) Whereas LAR is unifo rmly expressed throughout the non invasive tumor, (D) it is entirely absent from diffusely invasive lesions. (n = 6/group) Scale bar = 50 m
92 Figure 2 6. Co transplantation of invasive and non invasive tumors sequesters the invasive tumor within the b oundaries on the non invasive tumor mass. (A) Panoramic fluorescence micrograph of co transplanted U 87MG with firefly luciferase (f Luc) expressing hGBM L19 illustrates the restriction of invasion by entrapment within the confines of a non invasive lesio Luc labeled hNestin expressing tumor mass. Invasive cells approach the edge of the network of a strocyte fibers that circumscribe the lesion. (n = 9) Scale bar = 50 m
93 Figure 2 7. Human clinical specimens recapitulate the inverse relationship between CSPG expression and diffuse invasion. (A) The grade II glioma, pleomorphic xanthoastrocytom a (PXA) typically presents as a discrete lesion with well defined borders. (B) CSPGs are robustly expressed throughout the lesion and precisely define the edge of the tumor mass. (C, D) In contrast, the highly infiltrative grade IV glioma, GBM is largely devoid of glycosylated CSPGs aside from minimal staining specifically at sites of pseudo palisading necrosis.
94 Figure 2 8. Putative model of CSPG mediated invasion inhibition. When present, CSPGs induce dramatic rearrangements in the astrocyte and microglial populations associated with the tumor. These tumor cell glial cell interactions, mediated through the tumor ECM, define the tumor microenvironment and presumably, participate actively and directly in the regulation of the invasive character of a brain lesion.
95 CHAPTER 3 ZEB1 MEDIATES INVASI ON AND CHEMORESISTAN CE OF GLIOBLASTOMA Despite intense research efforts, glioblastoma remains one of the most lethal types of cancer. In particular, tumor recurrence after surgical resection and radiation f requently occurs regardless of aggressive chemotherapy. This recurrence has been attributed to residual cancer cells that can re initiate tumor growth. Here, we provide evidence that a subpopulation of highly invasive and chemoresistant glioblastoma cells is maintained by the transcription factor ZEB1 (zinc finger E box binding homeobox 1). ZEB1 is preferentially expressed in invasive glioblastoma cells, and its knockdown results in a dramatic reduction of tumor invasion as well as increased sensitivity to the chemotherapeutic agent Temozolomide (Temodar, TMZ) in vitro and in vivo. We find that this activator of epithelial mesenchymal transition (EMT) controls expression of the chemoresistance mediating enzyme MGMT (O 6 Methylguanine DNA Methyltransferase ) through the transcription factor c Myb, as well as cell cell adhesion pathways, thus linking chemoresistance and brain tumor invasion. Moreover, ZEB1 expression in glioblastoma patients correlates with tumor grade and is predictive of shorter survival. These results indicate that invasive glioblastoma cells are particularly sheltered from current therapeutic approaches, rendering them likely candidates for tumor recurrence. We thus identify ZEB1 as an important candidate molecule for glioblastoma recur rence, a potential marker of invasive tumor cells and a promising therapeutic target. Results With a low median survival of about 15 months, glioblastoma is the most frequent and most aggressive of all gliomas. A hallmark of glioblastomas is their high pr opensity
96 to invade the surrounding parenchyma, where single invasive tumor cells can be found far away from the primary site and frequently cross into the contralateral hemisphere. These cells cannot be isolated for surgical resection, or targeted by irrad iation. The current standard of care therefore includes chemotherapy to ablate invasive cells. Here, we provide evidence for a subpopulation of tumor cells that invade as single cells and demonstrate enhanced resistance to the chemotherapeutic agent Temoz olomide. Combined with their higher resistance to chemotherapy, these invasive cells leave the primary tumor site, invade deeply into the surrounding parenchyma, and thus evade all standard therapeutic regimens, rendering them likely candidates for gliobla stoma recurrence. We further identify the EMT activator ZEB1 as a master regulator of both invasion and chemoresistance pathways. In solid tissue tumors outside the brain, ZEB1 initiates distant spread and metastasis, and while gliomas rarely metastasize, an EMT like conversion has been speculated as a mechanism for their invasive phenotype. ZEB1 is a transcriptional repressor of cell adhesion molecules, such as E cadherin, miRNAs particularly the miR 200 family and cell polarity associated genes, and has emerged as one of the master regulators for mesenchymal transition and metastasis, and plays a critical role in the initiation of distant tumors. We found that ZEB1 expression confers chemoresistance, as well as invasiveness, thus generating a highly m alignant phenotype that may represent the basis for tumor recurrence. This is reflected in poorer clinical outcome of ZEB1 expressing tumors. Isolation and Characterization of Invasive Tumor Initiating Cells To better understand the cells responsible for i nvasion and treatment resistance, we postulated that invasive tumor cells are characterized by a lower proliferation index than other cancer cells and employed a paradigm that allowed us to separate cell
97 populations based on their proliferation rate. Prim ary tumor sphere cultures were loaded with the amine reactive fluorescent dye carboxyflurescein succinimidyl ester (CFSE) that is diluted equally to the daughter cells. This enabled separation of high proliferating (HP) and low proliferating (LP) cell frac tions based on fluorescence intensity proportional to dye dilution (Figure 3 1A and 3 5). We tested this hypothesis in an in vitro scratch assay and found that LP cells were more migratory than HP cells (Figure 3 1B). Next, we transplanted both cell popu lations into the striata of immunocompromised animals, where both LP and HP cells gave rise to fatal tumors with a median survival time of 153 and 177 days, respectively. Concomitant with LP cells being more migratory in vitro, we observed that LP cells g ave rise to significantly more invasive tumors in vivo relative to the HP cohort (Figure 3 1C and 3 6). Reduced cell cell adhesion in glioma correlates with invasiveness and N cadherin expression is proportional to the proliferative index of gliomas. Immu nohistochemistry revealed expression differences of cell adhesion molecules between LP and HP derived tumors. While HP cells showed high levels of membrane associated N cadherin and beta catenin, no staining was found in LP cells (Figures 3 1D and 3 2). W e concluded that loss of cell cell contact facilitates invasion of LP cells. Immunoreactivity to ZEB1, known to regulate expression of cadherins in cancer was absent in HP derived tumors, but prominent in LP derived tumors, especially in the invasion zone (Figure 3 1F, arrows) and in single invading cells distant from the primary tumor mass (Figure 3 1G). Knockdown of ZEB1 Results in Decreased Invasion and Increased Chemo Sensitivity We next tested whether LP cells are more resistant to the standard chemo therapeutic agent TMZ. We analyzed the susceptibility of LP and HP cells to TMZ
98 in a cell viability assay (Figure 3 2A), and found that LP cells displayed a significantly greater resistance to TMZ (concentration inhibiting viability by 90% (IC90) LP: 113 m M, IC90 HP: 1.1 mM, F test, p=0.0111). This suggests that cell cycle kinetics (i.e. in LP cells) provide a link between tumor invasion and chemoresistance, but at a higher level than expected from actual proliferation differences (Figure 3 2). This link wa s supported by the observation that TMZ resistant cells showed similar capacity for migration and invasion as LP cells (Figure 3 2B D). These data are substantiated by the clinical observation that tumors frequently recur despite aggressive chemotherapy, a nd establish a specific glioblastoma cell subpopulation that is more invasive and chemoresistant, which exclusively expresses ZEB1. Mechanism of ZEB1 Mediated Chemoresistance We therefore tested whether knocking down ZEB1 in tumor cultures results in red uced invasion and chemoresistance (Figure 3 3A F). ZEB1 knockdown cultures showed a prominent reduction in ZEB1 mRNA levels compared to control tumor cultures, both in sphere and adherent conditions (Figure 3 3A). ZEB1 knockdown was additionally accompan ied by reduced migration (Figure 3 2), reduced invasion (Figure 3 3C), and increased expression of membrane associated N cadherin and beta catenin (Figures 3 3D and 3 7). Cell viability analysis revealed that ZEB1 knockdown significantly increases the sen sitivity to TMZ (Figure 3 4A). We tested whether knocking down ZEB1 increased TMZ sensitivity of the most resistant population (i.e. LP) in vivo. Indeed, injection of therapeutic concentrations of TMZ into tumor bearing animals grafted with either shZEB1 L P or shGFP LP cells resulted in a reduction of tumor size in shZEB1, but not in shGFP animals (Figure 3 4B). TMZ had little or no effect on invasive
99 cells (shGFP), but prominently depleted the tumor mass (Figure 3 4B). Together, these data demonstrate that ZEB1 links tumor invasion and chemoresistance. The Complementary Nature of Invasion and Chemoresistance To elucidate the molecular pathways affected by ZEB1, we isolated proteins from adherent control and ZEB1 knockdown cultures and confirmed reduced expr ession levels of ZEB1 protein by western blot. N cadherin expression is inversely correlated with tumor invasion, but we found no significant change in total N cadherin protein levels (Figure 3 4C). Interestingly, rather than the expected reduction in tot al protein level, we found that the distribution of N cadherin had changed after ZEB1 knockdown. Specifically, ZEB1 knockdown induced an enrichment of N cadherin within the cell membrane, therefore enabling N cadherin mediated cell cell attachment and ret arded cellular motility (Figure 3 2). N cadherin is connected to the cytoskeleton by alpha and beta catenin, and the axon guidance molecule ROBO1 (roundabout homolog 1) can sever this connection. ROBO1, though not its ligand Slit, is expressed in malignan t glioma, and the ROBO1 sequence contains 3 binding sites for miR 203 20 (Figure 3 9). We showed that ROBO1 protein expression levels are reduced in ZEB1 knockdown cells, supporting the hypothesis that with active ZEB1, ROBO is expressed and disconnects N cadherin from the cytoskeleton, which enables ZEB1 expressing cells to shed cell cell contacts and invade as single cells (Figure 3 4C). The enzyme MGMT (O 6 Methylguanine DNA Methyltransferase 21) confers resistance to TMZ, and MGMT expression in glioma is prognostic for chemotherapeutic success. We observed reduced expression of MGMT in ZEB1 knockdown cells, concomitant with increased sensitivity to TMZ in these cells (Figure 3 4C). The MGMT sequence does not contain binding sites for known ZEB1 regulat ed microRNAs, nor
100 does the MGMT promoter contain sequences that would predict direct regulation of MGMT expression by ZEB1. However, the MGMT promoter contains predicted c MYB regulatory elements, and one study found c MYB associated with the MGMT promoter in breast cancer cells. c MYB contains at least five predicted binding sites for miR 200c and miR 203 20 (Figure 3 9). Importantly, ZEB1 knockdown reduced expression of c MYB (Figure 3 4C), supporting the hypothesis that MGMT expression is mediated by ZEB 1 through its indirect regulation of c MYB (Figure 3 4D). We explored this by modulating expression of the intermediate regulatory elements in this pathway, i.e. miR 200 and c MYB. Induced expression of miR 200c resulted in decreased expression of c MYB, M GMT and increased chemosensitivity, while antagonizing miR 200c resulted in increased expression of ZEB1, c MYB, MGMT, and increased chemoresistance (Figure 3 4E). Knocking down c MYB reduced, while forced expression increased, MGMT expression and chemores istance, respectively, but had no effect on ZEB1 expression (Figure 3 4E). Expression of c MYB in ZEB1 knockdown cells restored MGMT expression in these cells and rescued chemoresistance (Figure 3 4E). Our data identify ZEB1 as regulator of malignant glio ma invasion, which is further corroborated by immunohistochemistry and immunoblotting in specimens of invasive and non invasive tumors (Figure 3 10). Strikingly, ZEB1 expression correlated significantly with MGMT expression in glioblastoma samples (Figure 3 4C). Consequently, we found that ZEB1 expression also correlates with reduced survival of glioblastoma patients (median survival ZEB1+ 11.0 months (n=13), median survival ZEB1 19.2 months (n=20), Log rank test, p=0.0014, Figure 3 4). Furthermore, ZEB1 expression correlates with poorer outcomes in TMZ treated patients, which are reflected in both shorter survival and
101 shorter duration of successful TMZ therapy (Figure 3 8). These findings emphasize the relevance of ZEB1 for glioblastoma invasion and chem oresistance. Expression of ZEB1 is associated with migratory and invasive cells that are further characterized by reduced cell cycle kinetics (Figure 3 1) and TMZ resistance (Figure 3 3). Low proliferating cells have been identified and correlated with po or prognosis in a number of cancers and ZEB1 mediated invasion and chemoresistance may contribute significantly to this phenomenon. In human brain tumor specimens, ZEB1 expression is correlated with invasion and MGMT expression (Figure 3 4). Knockdown ex periments demonstrate the functional importance of ZEB1 in regulating invasion and chemoresistance and indicated that these processes are interrelated (Figure 3 8). ZEB1 stands at the center of an intricate pattern of direct and indirect regulatory mechan isms, which is executed through the miR 200/ 203 families (Figure 3 3). These microRNAs directly inhibit expression of the transcription factor c MYB, which in turn activates MGMT expression (Figure 3 3E). MGMT is a suicide enzyme, and two MGMT monomers ar e covalently linked upon its DNA repair activity. Of note, we find protein levels of monomeric, active MGMT to be very low in glioblastoma cells, while the majority of cellular MGMT exists in the exhausted, dimeric form (Figures 3 4E and 3 11). In fact, we were only able to observe monomeric MGMT in LP cells (Figure 3 4C). This indicates (a) high levels of DNA repair in glioblastoma cells and (b) that minor changes in MGMT expression may have pronounced therapeutic consequences. Loss of cell adhesion is ac hieved by disconnecting N cadherin from the cytoskeleton through ROBO1. ROBO1 expression may be regulated analogously by ZEB1 inhibiting miR 203 mediated interference. Hence, ZEB1 is a regulator of invasion and
102 chemoresistance in glioblastoma, a candidate agent for tumor recurrence, and may be a promising target for tumor therapy. Methods Cell Culture Tumor cell lines were generated (Piccirillo et al., 2006) and maintained (Siebzehnrubl et al., 2009) as described. Briefly, 50,000 cells were seeded per ml o f culture medium (N2, Invitrogen, Carlsbad, CA) in the presence of mitogens (20 ng/ml each of EGF and FGF2, Sigma, St. Louis, MO). Cells were propagated as spheres and passaged using Accutase (PAA, Clbe, Germany) every 7 days. For experiments with adhere nt cells, spheres were dissociated and plated in N2 medium supplemented with 1 % fetal bovine serum (FBS). For scratch assays, 2 x 106 cells were plated per well of a 6 well culture plate in N2 containing 1% FBS, and grown to confluence overnight. Confluen t monolayers were scratched with a pipette tip, and imaged at the time of the lesion and 24 hours later. TMZ resistant cells were obtained by treating a sphere culture with 800 M TMZ for 3 days. Medium was changed after that and surviving cells were grown up into a new culture. For sphere forming frequency assay, single cells were plated in 96 well cluster plates at a density of 1,000 cells/well. Three weeks after plating, the number of spheres > 50 m was counted. CFSE Loading and FACS Fluorescent dye lo ading and FACS sorting of low and high proliferating cancer cells were performed as described (Deleyrolle et al., 2011), with the exception that the lowest 5% of CFSE+ cells were used as HP. Briefly, cells were incubated with CFSE (Invitrogen) during regul ar passaging, washed three times and plated at 50,000 cells/ml. Seven days after loading, CFSEhi (LP) and CFSElo (HP) cells were sorted on a BD
103 FACSAria2 flow cytometer, collected in medium, counted and plated again at 50,000 cells/ml. These fractions were subcultured for 7 days before being used in subsequent in vitro or in vivo experiments. Cell Viability Assay The Methyltetrazolium bromide (MTT) assay was used as indicator of cell viability and performed as described (Holsken et al., 2006). Briefly, 10,0 00 cells were plated per well into 96 well cell culture plates and treated one hour after plating with varying concentrations of TMZ (ranging 5 M 5 mM, Tocris, Ellisville, MO). Five days after plating, cell viability was assessed by MTT dye conversion and analyzed on a plate reader (excitation wavelength 580 nm, reference wavelength 450 nm). Concentration effect curves for TMZ treatment were generated by nonlinear regression analysis as described 33. All curve slopes were larger than unity (F test, p>0. 05); therefore pIC50 and pIC90 values were obtained from curves with variable slopes. Knockdown Experiments Plasmids for knockdown of ZEB1 and expression of hsa miR 200c, as well as antago miR 200c and control sequences are described in (Brabletz and Brab letz, 2010). Plasmids for knockdown of c MYB were obtained from OpenBiosystems (Lafayette, CO). The plasmid for expression of human c MYB (Clarke et al., 1988) was a kind gift of Dr. J.S. Lipsick (Stanford University). Cancer cells were transfected using L ipofectamine selected using puromycin (Sigma) before being used for subsequent experiments. Animal Experiments Adult female Fox Chase SCID mice (Charles River, Wilmington MA) were used for in vivo tumor transplants. All procedures were performed according to NIH and
104 institutional guidelines for animal care and handling. After animals were deeply anaesthetized using USP grade Isoflurane (Halocarbon, North Augusta, SC), an incision was made in the scalp, the skull demonstrated and a hole drilled at the coordinates Bregma 0.5 mm anterior and 1.5 mm lateral. A Hamilton syringe was lowered 2.5 mm into the burr hole, and 1 l of a cell suspension was injected over 5 min before the needle was retracted. After the incision was closed with surgical staples the animal was allowed to recover before being returned to the cage. For survival and tumor end stage analysis, animals were transplanted with 10,000 cells, and tumor bearing an imals were scored regularly for tumor related symptoms. Moribund animals were anaesthetized and transcardially perfused with 4% paraformaldehyde in saline, the brains removed, postfixed and prepared for histology. For tumor invasion analysis, animals recei ved grafts of 100,000 cells, and were perfused 12 weeks post injection. For in vivo TMZ treatment, animals received orthotopic grafts of 100,000 shGFP LP or shZEB1 LP cells. 14 weeks after transplantation, tumor bearing animals were intraperitoneally injec ted with 20 mg/kg TMZ in saline (final DMSO concentration 25%). Animals received two injections per week, spaced 48 hours apart, for a total of six weeks. 48 hours after the last injection, animals were perfused. Immunohistochemistry and Immunocytochemistr y Immunostainings were performed as described (Siebzehnrubl et al., 2009; Zheng et al., 2006). A table of employed antibodies, suppliers and dilutions can be found in the supplementary information. All secondary antibodies were obtained from Invitrogen and diluted 1:500.
105 Image Acquisition and Data Analysis Low power fluorescent images were taken on a Leica DMLB epifluorescence microscope (Bannockburn, IL) equipped with a CCD camera (Spot Imaging Solutions, Sterling Heights, MI). To obtain full images of br ain sections, multiple gray scale images were acquired per section using Spot Advanced software (Spot Imaging Solutions) and merged into a full image and inverted into black on white images using Photoshop CS4 (Adobe Systems, San Jose, CA). Photomerged ima ges were imported into ImageJ and threshold levels were adjusted to distinguish tumor from background. Using the wand tool, all outlines of positively stained (black) tumor areas were outlined in each section and the perimeter and area were measured. The r atio of the squared perimeter distance over the area (P2/A) was calculated and used to compare invasive properties of different tumors. Since P2/A is a dimensionless number, the resulting figure is termed ative of a more dissociated tumor, whereas a lower invasion index represents a more spherical tumor. For in vivo TMZ treatment outcome, tumor area was used as indicator of overall tumor size. High power images were taken on an Olympus BX 81 DSU spinning di sc confocal microscope (Olympus, Center Valley, PA) and projection images of z stacks were generated using Slidebook (Olympus) software. RNA Isolation and Quantitative Real time PCR Total RNA was isolated from tumor sphere or adherent cultures using the RN easy quantified on a Nanodrop Spectrophotometer (Thermo, Wilmington, DE), and 1 g of total RNA was used for cDNA synthesis as described (Siebzehnrubl et al., 2009). 25 ng of cDNA were used for quantitative PCR using the SYBR green PCR master mix
106 (Applied Biosystems, Carlsbad, CA) on an ABI 7900HT (Applied Biosystems) as previously described 32. Expression levels of ZEB1 were quantified in triplicate relative to beta actin described elsewhere (Wellner et al., 2009). Protein Isolation and Western Blotting Proteins were isolated from adherent cancer cell cultures and primary tumor specimens obtained from the F lorida Center for Brain Tumor Research (FCBTR) as described (Siebzehnrubl et al., 2009). For western blotting, 20 g of denatured protein were loaded on 4 12% Bis Tris reducing gel (Invitrogen), separated and blotted onto a PVDF membrane (iBlot, Invitrogen ). Blots were blocked and probed with respective primary and secondary antibodies (see supplement) as described 32, and developed using the ECL Plus kit (Amersham, Piscataway, NJ) on a FuorChemQ Multi Image III (Cell Biosciences, Santa Clara, CA) and Alph aInnotech software version 126.96.36.199. Band densitometry was performed using ImageJ. Statistical Testing All statistical analyses were performed in GraphPad Prizm 5.0 (GraphPad Software, La Jolla, CA). Statistical tests are indicated in the text. In all analy ses, a P Pearson test for normal distribution of values. Observers were blinded to the patient data (including survival time) when performing the ZEB1 analysis.
107 Figure 3 1. Isolation and ch aracterization of invasive tumor initiating cells. Paradigm for the isolation of migratory (low proliferation, LP) and proliferative cells (high proliferation, HP). Top 5% CFSEhi (LP) and bottom 5% CFSElo (HP) fractions were separated 7 days after CFSE loa ding (A). In vitro scratch assays showed an increased capacity for migration in LP cells (B, mean +/ s.e.m., two tailed t test, p<0.0001, n=3). In vivo transplantation of LP and HP cells resulted in tumor formation with distinct invasive properties (C). Scale bars 10 m. Human specific nestin immunostaining reveals tumor invasion from LP cells (mean +/ s.e.m., two tailed t test, p<0.0001, n=5, a.u. arbitrary units). HP tumors stain for N cadherin and beta catenin, while LP tumors are negative (D). Scale bars 10 m. Fluorescence micrographs in D are projection images of confocal z stacks. While HP tumors are negative for ZEB1 (E), LP tumors show prominent reactivity at the invasion zone (arrowheads), which tapers off towards the tumor center (asterisk, F). Scale bars 20 m. Single invading LP cells prominently express ZEB1 (G). Scale bar 10 m.
108 Figure 3 2. Knockdown of ZEB1 results in decreased invasion and increased chemosensitivity. qPCR demonstrates upregulation of ZEB1 mRNA (A) during adhesion ( adh) of control cells, but prominent downregulation in ZEB1 knockdown experiments (ANOVA, p<0.0001, n=3). ZEB1 knockdown results in significantly reduced migration in vitro (B, mean +/ s.e.m., two way ANOVA, p<0.0001, n=3) as well as reduced invasion in vivo (C, mean +/ s.e.m., t test, p<0.0001, n=5, scale bars 10 m, a.u. arbitrary units). shZEB1 derived tumors are negative for ZEB1 (D), but show cell surface expression of beta catenin and N cadherin. Scale bars 10 m. Projection images of confocal z st acks.
109 Figure 3 3. Mechanism of ZEB1 mediated chemoresistance. MTT assays (A) show increased TMZ sensitivity in ZEB1 knockdown cells. IC50 shGFP: 152 M, IC50 shZEB1: 15 M, IC90 shGFP: 26 mM, IC90 shZEB1: 1.5 mM. IC50 and IC90 values differed signif icantly (F test, p=0.0011). ZEB1 knockdown and control tumor bearing animals were treated i.p. with either TMZ or vehicle (n=5 for each group, B). Fluorescence micrographs of representative tumors show that TMZ treatment affects the tumor mass (shZEB1), b ut not the invasive cell population (shGFP). Scale bars 100 m. Tumor size was reduced after treatment in the ZEB1 knockdown group, but not in the control group (two way ANOVA, p<0.0001). Dashed line represents vehicle treated tumor size used for normaliza tion. Western blots (C) show reduction of ZEB1 in shZEB compared to control cells, as well as reduced levels of MYB, MGMT, and ROBO1. GAPDH serves as loading control. A putative model (D) links ZEB1 to invasion and chemoresistance; both pathways are regula ted through intermediary factors (miR 200/ 203, ROBO1 and MYB, respectively). (E) Immunoblot analysis of c MYB and MGMT regulation. Forced expression of hsa miR 200c reduces, while antagonizing miR 200c increases, protein levels of ZEB1, c MYB and MGMT. Kn ockdown of c MYB reduces, while forced expression increases, expression of MGMT. Expression of c MYB in ZEB1 knockdown cells can rescue expression of MGMT and chemoresistance. Full length blots are presented in Supplementary Figure 6. Bar graphs depict rel ative cell viability of respective cell cultures exposed to 1 mM TMZ for 96 hours (normalized to DMSO treated control cells, n=8, p<0.0001, one way ANOVA with Bonferroni post test).
110 Figure 3 4. ZEB1 expression correlates with invasion in gliomas. No n invasive tumors show no immunoreactivity for ZEB1 (pilocytic astrocytoma WHO I), but invasive gliomas are strongly positive (glioblastoma WHO IV, A). Scale bars 20 m. Immunoblotting and densitometry demonstrate increased ZEB1 expression levels in gliobl astoma (GBM) over lower grade pilocytic astrocytoma (Pilo, two tailed t test, p=0.0473, B). In the glioblastoma cohort, expression of ZEB1 correlates with MGMT expression in these tumors (C, Pearson correlation, p=0.0022; depicted is linear regression and 95% confidence interval). Full length blots are presented in Supplementary Figure 7. Notably, not all glioblastomas express ZEB1 (red box), and ZEB1 negative tumors have a better prognosis (median survival ZEB1 positive 11.0 months, ZEB1 negative 19.2 mon ths, Log rank test, p=0.0014, D). In addition, survival intervals of TMZ treated ZEB1 patients were significantly increased (ZEB1+ 12.8 +/ 2.1 months, ZEB1 22.8 +/ 3.7 months, two tailed t test, p=0.0362), as was the average duration of TMZ treatment i n these patients (ZEB1+ 2.54 +/ 0.73 months, ZEB1 9.09 +/ 2.07 months, two tailed t test, p=0.0087). No significant age difference was found between the ZEB1+ and ZEB1 groups (average age ZEB1 positive 61.9 15.8 years, ZEB1 negative 58.1 11.7 year s, Mann Whitney test, p=0.15).
111 Figure 3 5. CFSE loading and FACS paradigm. Cells were loaded on day 0 with CFSE and the 5% CFSEhi and 5% CFSElo fractions separated based on CFSE intensity 7 days after loading. CFSE intensity inversely correlates wit h cell proliferation 12,30. Growth curves show differences in the expansion rate of LP and HP fractions, as well as these rates in relation to other paradigms employed in this study (unsorted, TMZR). Average expansion rates were 8.22 +/ 0.41 for unsorted cells, 9.25 +/ 0.82 for HP, 3.69 +/ 2.11 for LP, and 3.89 +/ 2.23 for TMZR. Bar graphs represent mean +/ s.e.m. A Kaplan Meier curve shows that transplantation of 10k cells/animal of both HP and LP cells results in fatal tumors. Tumor presence was conf irmed post mortem. Median survival HP: 153 days, LP: 177 days. Knockdown of ZEB1 does not affect proliferation rate significantly (two tailed t test, p=0.43).
112 Figure 3 6. Fluorescence micrographs and threshold images of representative tumors derived from LP, HP, shGFP LP and shZEB1 LP cells. Boxes in the images represent areas of larger magnification used in Fig. 1C and Fig. 3C, D. Scale bars 500 m. Fluorescence images were used to generate threshold images, which were used to measure invasion index (see methods). Bar graph demonstrates invasion index for a second tumor line with similar invasion patterns.
113 Figure 3 7. In vitro fluorescence micrographs of N cadherin and ZEB1 immunostaining. HP and shZEB1 cells show high membrane association of N cadherin, while LP and shGFP cells do not. Conversely, ZEB1 is absent in HP and shZEB1 cells, but present in LP and shGFP cells. Scale bars 20 m.
114 Figure 3 8. Complementary nature of invasion and chemoresistance. Sigmoidal curve fit of an MTT assay sh ows increased TMZ resistance in LP cells (A). IC50 LP: 53 M, IC50 HP: 13 M, IC90 LP: 113 mM, IC90 HP: 1.1 mM. IC50 and IC90 values differed significantly (F test, p=0.0111). TMZ resistant cells (TMZR) migrate similar to LP cells in vitro (B, two tailed t test, p=0.38). In vivo analysis of TMZR derived tumors reveals similar invasion to LP derived tumors (C, mean +/ s.e.m., n=4, two tailed t test, p=0.733). Scale bar 200 m. TMZR tumors are negative for N cadherin and beta catenin, but express ZEB1 (D). Scale bars 10 m. Fluorescence micrographs in D are projection images of confocal z stacks.
115 Figure 3 9. Gene regulation of c MYB and ROBO1. Shown are predicted 20 miR 200c/miR 203 binding sites in the UTR of the respective genes. The MGMT promoter se quence reveals predicted c MYB binding sites 23.
116 Figure 3 10. Full images of all western blots presented in Figure 3 3 Frames indicate excised bands. Arrows indicate monomeric (21 kDa) and dimeric (42 kDa) MGMT.
117 Figure 3 11. Full images of all western blots presented in Figure 3 4. Frames indicate excised bands.
118 CHAPTER 4 SEGREGATION OF HUMAN BRAIN TUMOR INITIATING CELLS AND GENES Despite exceptional efforts invested by academic research and the pharmaceutical industry, the prognosis of many b rain tumor patients remains dismal. The recently introduced neoplastic stem cell theory may offer an intriguing explanation, as it is implied that a rare population of CSCs drives the maintenance, propagation and potentially also recurrence of some of the most malignant tumors. This hypothesis contrasts to the stochastic model that attributes the capability of self renewal and continual growth to many of the tumor cells. Both models rely on the activity of clonal cell populations, and therefore, the most re asonable approach to study cancer would be the investigation of individual cells. However, there is a surprising lack of enabling is rather centered on the study of het erogeneous mixtures of cells from individual tumor samples, or alternatively, on the investigation of populations of enriched target cells expressing predicted traits (e.g. the CD 133 antigen). Here, we introduce a multi step procedure tailored to facilit ate (i) the identification of pertinent brain tumor samples, (ii) the unbiased separation of target cells, as well as (iii) their clonal propagation under adhesive culture conditions for comparative analysis and (iv) segregation of tumor initiating cells a nd genes. Our data demonstrate that various CSC phenotypes can be present in one tissue specimen, and furthermore, that their respective tumorigenic profile cannot be predicted from analyzing the heterogeneous mixture of cells in tumor samples.
119 Results Ide ntification of Pertinent Brain Tumor Samples Due to their infrequency, not every brain tumor biopsy may contain sufficient amounts of target cells, i.e. clonogenic cells with tumor sustaining potential. To first identify tissue samples harboring clonally active populations, we applied a semi solid, methylcellulose based neurosphere assay (mcNSA) that is also suitable for quantification of neural stem/progenitor cells (see Methods). This assay was applied to specimens obtained from 19 pediatric and one ad ult neurosurgery patients in order to determine clonal cell frequency. Clonal cell frequency represents the number of cells within a heterogeneous population, capable of generating a clonal sphere. Sphere forming cells are unquestionably rare. Depending on the particular tumor specimen examined, only 0.01 to 1.78% had sufficient proliferative potential to generate a clone. (Figure 4 1A, B). Yet, frequency or morphology of primary neurospheres did not predict multipotency and self renewal attributes. The se properties were found in only two cases, an anaplastic ependymoma (AE, #018) and a glioblastoma multiforme (GBM, #019). In both cases, multipotent, long term self renewing sphere forming cells continued to develop neurospheres for at least one year in c ulture (Figure 4 1C). Human brain tumor research currently does not distinguish between long term and short term activity in stem cell populations, but differing degrees of self renewal are acknowledged in the hierarchy of hematopoietic stem cells. As ther e may be parallels to neuropoietic cells, this long term, self renewing population could represent the top of the multipotent stem/progenitor cell lineages, potentially also responsible for sustaining the growth of brain tumor tissue. In this light, tissue cases 018 and 019 were selected for further discriminative studies.
120 The mcNSA is impractical for use in generating sufficient numbers of cells efficiently enough for subsequent investigations. During the period of one year in culture (15/16 neurosphere stage, respectively), the fraction of SFCs increased just slowly from 0.24 to 1.21% and from 0.30 to 1.27% in cases 018 and 019, respectively (Figure 4 1D). Additionally, the average size of neurospheres generated after the fifth passage (three months in mcNSA conditions) remained stable (200 400 cells/neurosphere), suggesting that SFCs and their generated number of cellular progeny reached an assay bound equilibrium (Figure 4 1E). Because of its limited practicality, the mcNSA was used only as an initial screening tool to identify specimens of particular interest; those containing a long term self renewing population. All additional analyses of these specimens were performed either directly on the resected tissue itself, or on cells derived from our adhere nt explant culture system (described below). (Figure 4 2). Isolation and Expansion of Heterogeneous Target Cell Populations In addition to long term proliferative potential, the migratory/invasive character of certain brain cancers underscores their patho novel explant culture system that has enabled us to isolate and analyze these migratory, potentially cancer stem like cells. This system offers an unbiased mechanism by which tumor tissue may be separated and en riched into various subpopulations based on the cells migratory competence and preference to attach to particular substrates at the time of tissue extraction (Figure 4 4A). The procedure was performed in parallel on untreated plastic dishes as well as unde r the influence of various substrate coatings (Laminin/poly L ornithine, LPO; gelatine; fibronectin; Methods and10) in analogy to earlier studies. Following a suitable time in culture (typically 7 10 days), the tissue explants were removed from the dishes, leaving behind
121 the separated, substrate specific migratory cell populations (mig). The remainder tissue pieces were dissociated into a single cell suspension and propagated as a separate population in their specific substrate condition (dis). Consequently up to eight distinguishable cell populations were available from each tissue specimen (Figures 4 2 and 4 3). Comparative analysis revealed highest frequencies of clonogenic, multipotent cells in LPOmig cultures, which were therefore were selected for fur ther analysis (Figure 4 4B). Both, 018 and 019 LPOmig cultures were characterized by heterogeneous cell morphologies present throughout the expansion period lasting for at least 35 population doublings (Figure 4 4C). Their respective cellular expansion rat es exceeded the rate of newly generated cells in the mcNSA by far. After 1302 days in culture, the total number of 018LPOmig cells increased by a factor of 2.68E+06, and for 019LPOmig cells by a factor of 2.37E+10. In comparison, the total number of mcNSA cells after the same period of time was reduced in case 018 by a factor of 4.86, and in case 019 by a factor of 1.25. Thus, serially passaged adherent cells rather than their sphere forming counterparts were selected for further analysis. To demonstrate t he suitability of stably expandable LPOmig cultures for the study of target cells, we used passage 10 cells for molecular profiling and investigation of tumorigenic traits. Surprisingly, both 018 and 019 LPOmig populations revealed characteristic sets of n eural stem/progenitor cell transcripts, (Figure 4 4D), but only 019LPOmig cells were tumorigenic. Not one of the 018LPOmig orthotopic xenograft experiments could recapitulate the features of an AE (n=0/8), while every engraftment of 019LPOmig cells develop ed into a tumor with GBM specific characteristics (n=4/4; Figures 4 4E, F and 4 5). The failure of 018LPOmig cells to induce AE features in our animal model may be due to low frequencies of tumorigenic
122 cells and/or potential mismatches of donor and host ce lls in the xenograft environment. Additionally, while little is known of the clonality of tumor sustaining AE cells, many recent reports predict these qualities as coinciding in GBM CSC populations. Thus, further analysis along the aims of our study was re stricted towards examination of 019LPOmig cells. Clonal Propagation and Comparative Analysis of Putative Target Cells To attribute the characteristics of multipotency, clonality, and tumorigenicity to individual cells and their progeny, heterogeneous 019 LPOmig cultures were seeded at ultra low densities in passage 5 (Figure 4 6A). As expected from the relatively low frequencies of SFCs found in the 019LPOmig cultures (0.39% at P+5, see Figure 4 4B), most adherent cells died under this condition within the first week. Yet, some individual cells survived, forming distinct colonies composed of uniform cells. From these, morphologically distinct colonies were selected, expanded for an additional five passages (P5+5), and subsequently evaluated in the mcNSA. Of the selected cell populations, two multipotent clones (CL6 and CL7) performed notably different from each other (Figure 4 6B, and 4 7). CL6 cells generated primary neurospheres at high frequency (16.4%) containing an average of 281 cells, while only 0.5% of the CL7 cells formed neurospheres with an average size of 99 cells. Because size and frequency of neurospheres could indicate biologically diverse subpopulations of stem like cells (Reynolds and Rietze, 2005), we next determined if CL6 and CL7 cells rep resented distinct CSC entities. Molecular profiling revealed a remarkable overlap of expressed genes, with an extensive list of stem/progenitor typic transcripts (Figure 4 6B, C). The exclusive expression of CD133 (Prominin 1) and the robust increase in MG MT expression (2.3 fold) within CL7 has provided two noteworthy contradictions. CD133 is
123 frequently used as a CSC indicator (Bao et al., 2006; Singh et al., 2004), and epigenetic silencing of the MGMT DNA repair gene (O6 methylguanine DNA methyltransferase ) by promotor methylation is associated with longer survival of GBM patients receiving alkylating chemotherapy. Nevertheless, not one CL7 xenograft (n=0/5) proliferated in situ, while every CL6 transplant (n=5/5) was tumorigenic in recipient NOD SCID mouse brains. Thus, our data suggests co existing stem cell phenotypes with distinct biological functions: CL6 as tumorigenic CD133 /MGMTlow cells, and CL7 as a non tumorigenic CD133+/MGMT+ cell population (Figure 4 6E). Analysis of transcripts involved in majo r pathways regulating the self renewal and the proliferation of stem and cancer cells alike (Pardal et al., 2003), revealed furthermore a variety of differentially expressed receptors, ligands, and target genes particularly in the Notch WNT p athways (Figure 4 6F). Compared to the profile of heterogeneous 019LPOmig cultures, patterns of differentially expressed CL6 and CL7 genes complemented each other, suggesting potential cellular interactions. This was further encouraged by preliminary in vi tro studies demonstrating an interdependence of CL6 and CL7 cells for maintaining their ability to self renew over prolonged periods of time (Figure 4 7). However, because these findings alone could not explain the discriminating tumorigenic potential of C L6 vs. CL7 cells, we were curious to evaluate their respective genomic configuration. Segregation of Tumorigenicity Genes Whole genome single nucleotide polymorphism (SNP) analysis was conducted on CL6 and CL7 cells (P5+5), using early and late stage 019LP Omig cultures (P+2 and P+10, respectively) for comparison. All samples showed GBM specific alterations20, including loss of heterozygosity (LOH) of tumor suppressor loci p53 and Rb1, as well as
124 amplification of the CDK6 and EGF receptor loci (not shown). S pecifically, examination of SNPs revealed a total of 34 LOHs that were identically present in all four investigated cultures (Figure 4 8A). This strongly related CL6 and CL7 cells to the GBM pedigree of case 019 and additionally justified the use of our in vitro expansion protocol. The tumorigenic CL6 cells, however, showed additional distinctive LOHs located on chromosomes 10p (39.1Mb), 11c (9.1Mb), 13q (0.7Mb), and 16p (6.4Mb). These LOHs were copy neutral (Figure 4 8B), potentially affecting a total of 4 82 gene products. Correlation with whole transcriptome data narrowed down the list of putative tumorigenicity genes to 81 that were actively expressed in CL6, CL7, and/or 019LPOmig cells (Figure 4 8C). 51 of these (63%) mapped to chromosome 10p, 27 (33%) t o 16p, 2 to 11c, and none to the 13q locus. Among the identified candidates, several take active part in the process of cancer development (i.e. C10orf7, CUL2, EIF3S8, FUS, NET1), (n=60/81; 74%), and showed a comparable expression level in CL6 vs. CL7 cell s. This finding corroborates that the four extra copy neutral LOH loci reflect a uniparental disomy, in which altered expression of affected genes depends on epigenetic control and complex transcriptome interactions. Yet, 21 genes from the extra loci were exclusively or differentially expressed in CL6 vs. CL7 cells, and one candidate suited to explain the tumorigenic behavior of CL6 cells. The Kruppel like transcription factor KLF6 (COPEB), a tumor suppressor gene frequently mutated in human prostate cancer was exclusively downregulated. Mutations of this gene are also observed in human GBM. More frequently, however, the expression of KLF6 may be affected by genomic losses on chromosome 10 that characteristically occur during gliomagenesis. Kimmelman et al recently demonstrated the functional relevance of KLF6 expression for
125 suppressing tumorigenic activity in human GBM cell lines (Kimmelman et al., 2004). In their work, the authors also reported highly variable degrees of KLF6 expression in primary GBM tiss ue. Similarly, our 019LPOmig cells showed high degrees of KLF6 expression (Fig. 4c), likely due to the activity of other cell phenotypes in the heterogeneous population that were not affected by the specific genomic aberration on 10p. Failure of KLF6 expre ssion in tumorigenic CL6 cells was therefore not predictable from heterogeneous 019 tumor cells. We conclude that for profiling of tumorigenicity cells and genes, the evaluation of individual tumor cells and their progeny is pivotal. Methods Cell Culture H andling and transfer to the lab of human tissue derived from consented patients and/or their guardians at the University of Florida Department Neurosurgery, was performed according to institutionally approved protocols. Brain tissue was minced under steril e conditions into chunks of ~1 mm3 volume (Figure 4 2) and randomly divided into two equal parts. One part was fixed in 4% paraformaldehyde (PFA) and stored until further use; the other vital tissue pieces were equally assorted to the mcNSA and the adhesiv e cell derivation protocols. Preparation of a single cell suspension for the mcNSA was conducted overnight in a 0.25% trypsin solution on a shaker at 4C followed by adding five percent fetal calf serum (FCS, HyClone, Logan, UT, USA) and gentle dissociati on using graded fire polished glass pipettes the next day. Trypan blue exclusion confirmed >95% vital cells in each case. Dissociated cells (100,000/ml) were distributed into non adhesive culture dishes in semi solid mcNSA media comprising: 1% mc; 5% FCS; modified N2
126 extract; 1X antibiotic antimycotic solution (abx); and 1,000 units/ml human LIF in DMEM/F 12. Epidermal growth factor (EGF) and basic fibroblast growth factor (bFGF), each at a concentration of 40 ng/ml, were added the first day, and at 20 ng/ml every other day thereafter. Unless otherwise specified, media and growth factors are available from Sigma (St. Louis, MO, USA), Invitrogen (Carlsbad, CA, USA), Millipore (Billerica, MA, USA) and/or R&D systems (Minneapolis, MN, USA). Cell culture plastic ware was purchased from Corning (Corning NY, USA). At week 2 3 in culture, primary neurospheres were counted (6 wells for each experiment), collected, centrifuged at Biosciences, San Jose, CA, USA) to ensure a single cell suspension. For derivation of higher degree neurospheres, the cells were re distributed in non adhesive culture dishes at densities of 50,000 cells/ml as described above. This passaging protocol was conducted for each case at intervals of 2 3 weeks, for as long as formation of neurospheres was observed (Figure 4 1). For analysis of multipotency, third degree neurospheres derived from the mcNSA were attached to Laminin/poly L ornithine (LPO) coated glass coverslips and maintained without growth factors for 14 35 days in Neurobasal medium containing 1x B27 supplement, 2mM L glutamine and abx. Cells were fixed in ice cold 4% PFA for 20 minutes. The mcNSA protocol was also applied to a variety of adherent cell populations at passages 5, 10, and 20 as specified in the results. For de rivation of adhesive cell populations (Figure 4 2), the remaining tissue chunks were further teased into the smallest possible fragments using a scalpel and
127 forceps. Approximately 50 micro fragments were placed per 6cm culture dish, in individual drops (50 addition to standard, uncoated plastic dishes, three defined surface coatings (i.e., LPO, FN, and GL; see10 for dish coating protocols) were used for the simultaneous preparation of adherent c ultures from each tissue case. During the course of one week, the volume of media around the adherent micro fragments was gently increased to a total of 4ml/6cm dish, by daily additions of proliferative media (P media) consisting of: DMEM/F12 supplemented with 5% FCS; modified N2 components; 1.1 x B27 EGF and bFGF (each 40 ng/ml) were added the first day, and 20 ng/ml of each were added every other day thereafter. After 2 10 days, cells were observed to migrate out and proliferate around the vicinity of the adherent tissue (Figure 4 4A). At this time, all tissue fragments (containing cells that were unable to migrate out onto the respective culture dish surface) were remove d, trypsinized, and distributed in P media into 6 cm dishes coated with the same substrate. Thus, for each adhesive condition, two types of culture preparations were made one containing migratory active cells (mig), the other containing dissociated (diss ), migratory delayed, resident cells from the same tissue specimen (Figure 4 3). P media was changed every four days; 1 g/ml laminin was continuously present in LPO conditions. For banking of adherent populations, cells from all conditions were grown to c onfluency and frozen without further passaging (P+0) in two cryotubes / 6cm culture dish (2.5 5x105 cells/cryotube). For preparation of clonal sub populations, adhesive cultures (P+5) were distributed in P media at densities of 2 20 cells/cm2 in LPO coa ted 10cm dishes. 20
128 individual locations of adherent cells were circled with a pen on the bottom of the dish at 3 days after plating (Figure 4 6A) and regularly photodocumented from this time on. At days 30 60 after seeding, selected colonies were separate d using 8 mm cloning rings, trypsinized, transferred to a 6cm LPO coated culture dish, grown to confluency, and stored in liquid nitrogen (P5+1) until further use. Expansion of clones and analysis in the mcNSA assay was performed as described above for sub strate specific adhesive cell populations. The proliferative capacity of adhesive cell populations was assessed expansion in P media on 6cm dishes coated with the respective substrates. Cells were grown to confluency, trypsinized, counted, and passaged in ratios of 1:2 for up to 20 passages. log l) + X, where (n) is the final PD number at end of a given subculture; (UCY) is the cell yield at that point; (I) is the cell number used as inoculum to begin that subculture; and (X) is the doubling level of the inoculum used to initiate the subculture being quantified. Cells were photodocumented at least every fifth passage using a Leica (Bannockburn, IL, USA) DM IRB microscope equipped with a DFC 300F camera system. Immunocytochemistry. The basic immunolabeling buffer contained PBS, 10% FCS, and, for intracellular antigens, additionally 0.1% Triton X 100. After blocking nonspecific antibody activity for 20 min in 5% goat se Promega; GFAP, polyclonal rabbit, 1:400, DAKO; CNPase, monoclonal mouse, 1:250, Chemicon) were applied for 4 hours at room temperature. Antigens were visualized using corresponding secondary antibodies (Jackson ImmunoResearch, West Grove,
129 PA, USA; or Molecular Probes, Eugene, OR, USA). Cell nuclei were labeled for 10min DMLB upright microscope and images were capture d with a Spot RT Color CCD camera (Diagnostic Instruments, Sterling Heights, MI, USA). Xenograft Experiments. Animal experimentation was conducted according to institutional IACUC guidelines. Cells were trypsinized, concentrated to a density of 2 10 x 104 frontal cortex of adult (>90 days) immuno compromised NOD SCID mice (Charles River, Wilmington, MA, USA). The stereotactic coordinates were: 2.5/ 0.5/1.2 and 1.5/ 2/1.3 (mm depth, A P, lateral), respectively. Engrafted mice were monitored on a daily basis, and tumor related behavioral abnormalities were videotaped. Animals were sacrificed and perfused transcardially with 4% PFA. Brains were removed and stored in 2% PFA until further use. MRI data were obtained using an 11 T magnet from fixed brains samples placed in PBS for the imaging session. For histological analysis and standard H&E staining, brains were placed in 2% PFA containing 30% sucrose (v/v) overnight, se stored in cryoprotectant. Scoring of tumor formation was based on evaluating every fifth section of the engrafted brains for the typical appearance of human tumor cells in the DAPI s taining. Molecular Biology. Total cellular RNA was isolated from neurospheres and adherent cell samples Valencia, CA, USA). Sample amplification and cDNA synthesis was perfor med using
130 100ng of total RNA for hybridization to the HumanRef 8 BeadChip according to the manufacturer's instructions (Illumina, Inc., San Diego, CA, USA; http://www.illumina.com/pages.ilmn?ID=197). Arrays were scanned at the microarray core facility at t he Burnham Institute for Medical Research (La Jolla, CA, USA) with an Illumina Bead array Reader confocal scanner according to the manufacturer's instructions. Data processing, analysis, and generation of heat maps was conducted using Illumina BeadStudio s oftware (Gene expression module v. 3.2.32). Background and rank invariant methods were used for normalization. Expression intensities with p values below 0.01 in at least one of four samples were log10 transformed and normalized by z scoring. Transcripts w intensity values differed more than 2 fold. For genome wide single nucleotide polymorphisms (SNP) analysis, DNA was Samples we re investigated for 561,466 SNPs on the Illumina HumanHap550 BeadChip Illumina BeadStudio 3.1 software including the Genotyping 3.1.12 and Genome Viewer 3.1.4 modules. Del etions were identified by examination of Log R ratio and B allele frequency. Gene lists were extracted using SNPper Gene Finder (http://snpper.chip.org/bio/snpper enter gene). Multiple chromosomal regions showed characteristics of duplication events bas ed on the intensity ratios of the two alleles. However, in most of these regions, the overall signal intensity (log R ratio) was not exceptionally high. This might be explained by an elevated intensity average due to the fact that a large proportion of the genome is duplicated or indicates that only a
131 subpopulation of the cells contains the duplications. Thus, due to higher validity, only deletions (resulting in LOH regions) were included in the analysis.
132 Figure 4 1. lt SFCs in the mcNSA (a) Single exposed to mitogens in non adherent, low density growth conditions, formed multi cellular spherical aggregates (neurospheres) within 21 7 days (phase contrast, left panel). 1 neurospheres were dissociated to a single cell su spension and plated to evaluate the generation of higher degree neurospheres (phase contrast, medium panels). For evaluation of multipotency, 3 neurospheres were plated, outgrowing cells were allowed to differentiate for 21 7 days, and subsequently asse s sed for expression of contained SFCs (respective frequencies, right), but lt SFCs were found in only two cases (arrows). (c) Increase of neurosphere and total cell numbers were documented for one year in culture (up to the 16 neurospheres stage). Note declining cell and neurosphere numbers in sample 020 (a medulloblastoma). SFC fractions did not increase significantly after the 5 neurosphere culture stage (d), and average cel ls/neurosphere remained constant (e). AE, anaplastic ependymoma; DAPI, nuclear marker 4',6 diamidino 2 phenylindole; GBM, glioblastoma multiforme; INF, sample lost due to infection; n.d., not determined; SFCs, sphere forming cells. Scale bars: a (left, ph
133 Figure 4 2. Cell derivation protocol. (a) Surgical tissue specimens were minced into 1mm3 sized chunks, and randomly selected pieces were allocated to two isolation protocols conducted in para llel. (b) For generating neurospheres in the mcNSA, chunks were dissociated to a single cell suspension and cultured for 2 3 weeks as described in the Methods section. Higher degree neurospheres were prepared by dissociation to a single cell suspension an d re plating of cells into identical mcNSA conditions. This procedure was continued for as long as neurospheres formation could be observed. (c) For separation of migratory active cells, tissue chunks were plated onto substrate coated culture dishes and c (mig) cultures formed from cells that actively traveled away from the plated tissue chunk within a week. At that time, chunks were removed, dissociated, and the single cell suspension was plated onto simi lar substrate coated and diss cells were grown to confluency and frozen as case and substrate specific adherent cultures in P+0.
134 Figure 4 3. Distinguishable cell populat ions result from the adhesive isolation protocol. Phase contrast pictures showing substrate specific migratory (mig) and non migratory (dis) cells of cases #018 and 019 at P+4 (population doubling 4 7). Individual cultures could be distinguished based on m or phology and distinct Abbreviations: dis, migratory inactive (dissociated) subpopulation; FN, fibronectin; GL, gelatine; hr, doubling time of cell population in culture (hours); LPO laminin/poly L ornithine; mig, migratory active subpopulation; no outgrowth, the tissue chunks of case 018 did not adhere on plastic dishes; PL, plastic.
135 Figure 4 4. Explant cultures generate target cell populations. (a) The photograph depicts an isolated tissue chunk (case 019) at 3 days in culture with migratory active cells moving away from the tissue (inset). (b) Comparative analysis of the various substrate specific 018/019 mig and dis subpopulations in the mcNSA revealed that LPOmig cultures contained highest frequencies of SFCs (arrows). Their respective multipotency and continuing ability to from higher degree neurospheres could be demonstrated in all cases, with the exception of FNdiss and GLdis cultures (stars), using 3 neurospheres in a nalogy to Figure 1a. Shown are the results from analysis of P+5 cells; similar findings were recorded from 018/019 LPOmig cells at expanded passages 10, 20, and 30 (not shown). (c) A stable expansion kinetic was recorded from both, 018 and 019 LPOmig cultu res, with heterogeneous cellular morphologies present throughout the observation period (insets, phase contrast at P+10). (d) Transcript analysis of 018/019 LPOmig cultures at P+10 suggested a continuing presence of stem/progenitor cells. Note that CD133 ( PROM 1) transcripts were not present in 018LPOmig cultures. (e) Coronal sequence of FLAIR (left) and corresponding T2 weighted (right) MRI sections of a NOD SCID mouse brain at 77 days after engraftment of 25,000 018LPOmig cells (P+10) into the lateral ve ntricle. No sign of tumor formation can be noted. (f) In contrast, the macroscopic appearance and a sequence of T2 weighted coronal MRI sections of a recipient mouse brain at 39 days after frontal cortex (asterisk) transplantation of 25,000 019LPOmig (P+10 ) cells. GBM specific signs include typical T2 heterogeneity, a midline shift, and also a crossing of the m idline with far distance spread along the rostro caudal axis acutely isolated cell; dis, migratory inactive (dissociated) subpopulation; FN, fibronectin; GL, gelatine; LPO, lami nin/poly L ornithine; mig, migratory active subpopulation; NSA, neurospheres assay; SFC, (neuro ) sphere forming cell.
136 Figure 4 5. 019LPOmig cells at early stages post engraftment. (a) The H&E stained tissue section shows a graft located in the fimb ria, 23 days after transplantation of 25,000 donor cells aimed at the lateral ventricle. At this stage, dense clusters of proliferative active donor cells were observed, with individual tumor cells infiltrating the surrounding host tissue, and also moving considerable distance across the midline (boxed area magnified left). (b) Frequently, multinucleated tumor giant cells (left) were observed, which was a characteristic finding of tumor cells present in the original 019 GBM tissue (not shown). Further typic al traits of engrafted cells at this stage after engraftment included their accumulation around host capillaries (centre), and the presence of small necrotic areas within the tumor mass (right).
137 Figure 4 6. Generation of cell clones from 019LPOmig c ultures. (a) Phase contrast micrographs of heterogeneous 019LPOmig cells and monomorphic CL6 and CL7 cells after clonal derivation. (b) The graph highlights the dissimilar performance of CL6 and CL7 cells (arrowheads) in the mcNSA. For comparison, data acq uired from 019aic cells, as well as from early and late passage 019LPOmig cultures are presented. Two other clonal cell populations (not shown) performed similar to CL6 cells, and thus were not chosen for further analysis. (c) The graph clusters the number of genes expressed in CL6 and CL7 cells according to their respective ratio. Note that the expression of most genes overlap. (d) Expression profiling of CL6 and CL7 cells revealed the presence of stem/progenitor typic transcripts in both populations. Cult ures were investigated at five expansion steps after cloning (P5+5). Note the specific expression of CD133 (PROM1) and nestin in CL7 and CL6 cells, respectively (asterisks). (e) Table summarizing the constellation of most relevant clinical characteristics of CL6 and CL7 cells. (f) Comparative analysis of differentially expressed genes in proliferation and self renewal pathways (asterisks indicate receptors, ligands, and target
138 Figure 4 7. Preliminary studies on interde pendence for long term self renewal of CL6 and CL7 cells. (a) Growth chart demonstrating stable expansion kinetics of CL6 and CL7 cells. The arrows indicate the culture stage cells were derived from for analysis in this study. (b) Phase contrast images (le ft) of morphologically distinguishable CL6 and CL7 neurospheres. In analogy to the procedure for aic cells, CL6 and CL7 neurospheres could be serially passaged in the mcNSA, and multipotency analysis was performed on 3 neurospheres. The respective 3 neu rospheres generated cells of mostly tubulin and/or GFAP). (c) The failure of CL6 and CL7 cells to self renew for prolonged periods of time was observed in the mcNSA. Both populations seized to generate neurospheres beyond the 5 culture stage. However, when CL6 and CL7 cells were combined, neurosphere formation continued.
139 Figure 4 8. Segregation of tumorigeniciy genes. (a) Graphic representation of all chromosomes and LOH loci r ecorded in the four investigated cell samples (chromosomal ideograms from Beadstudio software, Illumina). Note the overlap of 34 LOHs in all samples, as well as CL6 specific aberrations on chromosomes 10p, 13q, and 16p. The LOH on chromosome 11c was also e vident in expanded 019LPOmig cells, suggesting a relative increase of CL6 cells over time. (b) Sample file of chromosome 10 SNP analysis. The CL6 selective 39.1Mb large LOH locus is highlighted in red. Note the copy neutral LOH typic unchanged gene dose in the log R ratio graph (for comparison, an apparent deletion on 10q is highlighted; arrow). (c) Comparative analysis of the 81 expressed genes from the four extra LOH loci. Of the total 482 potentially affected gene products, the array recognized 225; 144 of these were neither expressed in CL6 or CL7 cells. In CL6 cells, three genes (AKR1C1, AKR1C2, CUGBP2) were exclusively expressed; two genes (AKR1C3, PTPLA) were differentially upregulated; 10 genes (BMP1, PFKP, CLN3, RSU1, STX4A, COMMD3, NMT2, MAPK3, DNA JC1, VIM) were differentially downregulated, and 6 genes (CAMK1D, BAMBI, COPEB, ZNF25, OPTN, MGC3121) were exclusively not expressed. The asterisk demarcates COPEB (KLF6).
140 CHAPTER 5 SYNTHESIS Our laboratory has developed a series of novel human GBM cell lines, which distance infiltration throughout the host brain upon intracranial transplantation. Further, the patterns of this microscopic insinuation have remained true to the ha llmark corridors of invasion reflected in the human disease the so called Secondary Structures of Scherer peri vascular satellitosis, subpial spread within and inferior to the meninges, and intrafascicular spread within patterned white matter tracts. With this study we demonstrate, for the first time, that high grade glioma infiltration into the brain occurs in the absence of a CSPG rich, inhibitory matrix. In contrast, lesions that strictly grow expansively, such as metastases to the brain, certain l ow grade gliomas, and the standard human glioma cell lines are characterized by a robust, CSPG containing ECM the tumor. Importantly, our findings suggest several, str iking departures from the canon of common knowledge for glioma cell invasion. The role of reactive astrocytes, the association between activated microglia and non invasive tumor types, and the absence of MMP mediated proteolysis are among the topics in wh ich our findings deviate from the widely accepted models of glioma cell invasion. In total, this finding support the assertion that tumor associated CSPGs are sufficient to induce the massive glial reorganizations that characterize the non invasive tumor microenvironment and the absence of these elements provides favorable conditions for the diffuse infiltration that typifies high grade glioma and GBM.
141 Aside from brevican, relatively little is known about CSPGs and glioma invasion. The work that has bee n done in this area has focused specifically on the activity of the brevican core protein, not the CS GAG side chains under scrutiny here. Convincing evidence has been brought to light associating diffuse infiltration with the presence of a variant of the bre vican protein. This unique brevican derivative, referred to either as the hyaluronan isoform (Viapiano et al., 2005) has been identified across an array of primary human glioma specimens and is entirely absent from non invasive brain lesions. These findings may appear at first to controvert our data. They have, in essence presented a credible link between glioma invasion and the presence of a CSPG, whereas we have clearly stated the oppo site. However, when we consider the details, our separate investigations are in fact, quite complementary. From our position, staining with the CS 56 antibody or WFA lectin reveals the GAG side chains rather than the core proteins, or variants of the cor e proteins themselves. Therefore, our work identifies these side isoform (or the HABD fragment as it was referred to originally) described by Hockfield and Matthews is clearly present and enriched in high grade human glioma, these authors go further to demonstrate that this molecule is actually an under glycosylated, or perhaps incompletely glycosylated variant of the full length brevican molecule. (Viapiano et al., 20 05) Thus, in both situations, diffuse infiltration corresponds to the relative absence of the otherwise inhibitory CS GAG side chains. Moreover, using our methods, although presumably present in abundance, this unique under glycosylated variant would lik ely go undetected. When taken together, rather than contradict one
14 2 another, our separate investigations have elucidated distinct and opposing roles for the proteoglycan core proteins and their CS GAG side chain in the context of glioma invasion. Astrocy tes have been extensively researched in the context of tumor origination. Persuasive evidence indicates that, for at least certain subclasses of glioma, reactive or stem like astrocytes may be the original tumorigenic seeds, initiating further tumor growt h. (Scherer, 1940; Silver and Steindler, 2009; Hambardzumyan et al., 2011) Beyond these cell of origin studies however, there is very limited information regarding their role in the microenvironment of an established tumor and even less that speaks to th eir specific contributions to the invasive niche. Other than the work presented here, Edwards and colleagues has offered what may be the only other examination of human glioma infiltration and reactive astroglia in vivo. (Edwards et al., 2011) These inv estigators describe a molecular interaction between reactive astrocytes and an adjacent glioma cell mass, which culminates in the reactive astrocyte population inducing tumor cell dispersal. Our work suggests that the situation described by Edwards and co lleagues may represent only one of potentially many reactive astrocyte tumor cell interactions that ultimately regulate brain tumor invasion. We have demonstrated that reactive astrocytes may either nurture or constrain tumor cell infiltration depending o n the presence of CSPGs and the corresponding state of the astrocyte population. On one hand, reactive astrocytes may foster microscopic tumor cell invasion presumably employing the connective tissue growth factor (CTGF) mediated mechanism detailed by E dwards and colleagues. On the other hand, in the presence of a robust, CSPG containing matrix as in the case of our co transplantation
143 experiments reactive astrocytes actually have the capacity to forcibly confine an otherwise invasive tumor cell popu lation. Thus, our data suggests that the reactive astrocyte brain tumor relationship is dynamic, allowing for both invasive and anti invasive modes, modulated through the tumor ECM. Additionally, these findings present the intriguing possibility that the tumor confining abilities of CSPG stimulated reactive astrocytes might be harnessed in a unique and novel approach toward therapeutic intervention against glioma cell dissemination. Analogous to the situation with astrocytes, common knowledge informs us that microglia facilitate tumor cell dispersal and contribute to a pro invasive microenvironment. (Charles et al., 2011) This work has largely been established with th e involvement of matrix metalloproteinase 14 (MMP 14 or MT1 MMP) bound to the membrane of tumor associated microglia. These authors also suggest that the microglial derived MMP 14 may activate glioma derived pro MMP 2 into active MMP 2, thereby adding to the invasive capacity of the tumor. (Markovic et al., 2009) Our data reaffirms the association between microglia and invasion, and also establishes a potentially novel concurrence of microglia with non invasive tumor growth. To our knowledge, there has been no definitive evidence presented that speaks directly to microglial inhibition of invasion. However, because they are clearly present and active in the case of invasion, we reason that their predominance, coupled with their alternate morphology sugge sts that they may also actively participate in the pathogenesis of non invasive tumors. Strikingly, our work provides an explanation for this dramatic morphological and putatively functional difference. Because the composition of the
144 ECM is so robustly d ifferent between these separate tumor types, we contend that these two, distinct cell states arise as a consequence of the microglial matrix interaction. A conclusion supported by our in vitro microglial activation assay. The presence of tumor associated CSPGs triggers microglial activation, whereas their absence preserves the microglia in a state of rest. Thus, a synthesis of our work with the preponderance of data that suggests that microglia are pro invasive, suggests that the microglia tumor interact ion is not fixed. Rather, microglia may either foster or potentially inhibit tumor invasion depending on their specific activation state, which in turn is regulated through the tumor ECM. The association between CSPGs and non invasion is something of a paradox when considered from the standpoint of developmental biology. During neural development, CSPGs serve as inhibitory guidance molecules, delineating boundaries between functionally different emerging brain structures. (Cooper and Steindler, 1986; S teindler et al., 1988; Snow et al., 1990) Essentially, these matrix rich outlines inhibit or even repel axon outgrowth away from a proteoglycan barrier. Therefore, it is something of a paradox that non invasive tumor cells remain within an environment th at is so rich with these classically inhibitory guidance molecules. By extension, when we suppress environment that would allow cells to remain in place, yet under these conditi ons the tumor cells disperse. Our observations with the LAR phosphatase receptor offer a uniquely simple explanation for these otherwise curious results. Essentially, the non invasive phenotype is an active, receptor mediated pathology, in which the tumo r cells anchor themselves within the fibrils of a CSPG rich ECM. The LAR receptor acts as a
145 tether, securing the cells to their secreted matrix and in turn preventing the matrix molecules from diffusion. In consequence, the non invasive tumor develops as a precisely focused concentration of proteoglycan. In addition, the absence of LAR from the reactive astrocyte population helps to clarify their unique response to the developing lesion. Recall that the LAR staining pattern precisely defined the non inv asive tumor itself and was not seen in the surrounding reactive astrocytes. Thus, whereas the LAR expressing tumor cells are held in place, the reactive astrocytes are potently repelled away to the boundaries of the CSPG rich field of matrix. Importantly this scheme is removes the ligand for the LAR phosphatase receptor. Analogous to a function blocking antibody or peptide, the absence of the ligand compromises receptor me diated coupling to the proteoglycan rich matrix. Consequently, the constraints on this otherwise non invasive tumor cell population are lifted and we observe microscopic dissemination of the tumor. The concept that glioma cell invasion might occur in th e absence of a CSPG rich inhibitory matrix is highly controversial. Numerous studies have indicated that the breakdown of an initially present inhibitory ECM is a required first step before invasion is possible. (Rao, 2003) Although this is clearly the case for epithelial tumors, our data suggests that CSPG mediated inhibition of invasion may represent a minimal, or perhaps negligible barrier, from the very earliest stages of high grade glioma development. However, we were concerned that diffusely infil trative lesions may have initially produced CSPGs, only to be immediately proteolyzed by abundant and highly active proteases. Although our data did not support this notion no evidence for
146 functional MMP 2 or 9 was detected in the invasive tumors we c annot rule out the possibility that other proteolytic enzymes may contribute to the diffusely invasive phenotype and we do not intend to discount protease mediated tumor invasion in general. Rather, our findings substantiate a need to place CSPGs, and mat rix proteolysis in context within the overall tumor pathology. For instance, CSPGs are observable, albeit at relatively low levels, within and surrounding sites of pseudo palisading necrosis both within our animal models as well as human pathological spec imens of glioma. Thus, these matrix degrading mechanisms may apply to these specific zones, however may have little or no applicability to the tumor in general, and especially to the invasive edge where CSPGs are essentially absent. In this light, our tw o distinct tumor types may better represent the opposite ends of a spectrum of invasion comprising CSPG rich, expansive lesions on one side, and CSPG absent, diffusely infiltrative tumors on the other all parts of which may be present within a single, complete glioma. In total, (Figure 2 8) we have presented the novel finding that the relative presence or absence of micro environmental CSPGs correlates with the invasive character of human glioma. Using primary human glioma cell lines as well as patho logical human specimens, we have demonstrated that high grade, diffusely infiltrating lesions are associated with the relative absence of CSPGs, whereas their robust presence is associated with strictly expansive tumor growth. Further, the strong presence of these potently inhibitory molecules within the non invasive tumor microenvironment induces sweeping morphological and/or architectural changes in tumor associated astrocytes and microglia that may contribute to the active containment of the tumor cell population.
147 Conversely, the absence of these tumor glial cell interactions preserves an environment that is uniquely conducive to widespread tumor cell insinuation. Importantly, amplifying or attenuating the levels of CSPGs within the tumor microenvironm ent can directly modulate tumor cell infiltration. This work presents a novel insight into the basic biology of the brain tumor microenvironment, the ECM, and their synergist influence over microscopic tumor cell invasion. The findings from Chapter 4 ad dress a series of considerations prevailing in the field of cancer research. First, the phenotypic and molecular profiles of rare cell populations that drive aggressive types of cancer are not necessarily apparent among the heterogeneous cells of a tumor s ample. Second, a distinctive profile of tumorigenic cells remains difficult to predict based on existing theories; and third, more than one clonal cell population may be responsible for tumor maintenance. In contrast to prospective isolation methods that aim for purification of target cells with predicted qualities, such as the expression of specific cell surface markers or the exclusion of Hoechst dye by side populations, we initiated our experiments with a heterogeneous population. Because extensive self renewal is regarded as a cardinal feature of stem cells and CSCs alike, analysis was restricted to tissue samples that showed a profound content of multipotent, clonally active, long term self renewing cells, (as determined in the mcNSA). The rationale of this approach was to access the highest possible degree of phenotypic diversity present among tumor residing cells at the time of surgery for expansion and retrospective analysis of the tumor sustaining cell fraction. To facilitate cell extraction from tu mor tissue without bias, we offered a spectrum of culture substrates that attracted diverse subpopulations of tumor cells from tissue
148 explants, additionally enabling the separation of local migratory from non migratory cell populations. Cells derived from this system were rapidly expandable as heterogeneous or as single cell derived adherent cultures without losing the characteristic genomic traits of the initial starter cell population. Thus, a plentitude of material was made readily available that enable d the reliable evaluation of the tumorigenic and stem/progenitor qualities of the original specimen. The successful use of this approach was exemplified on two individual, single cell derived populations isolated from GBM tissue. Clonal activity was a ra re feature among 019 GBM cells, however, characterization of two clonally expandable cancer stem like clones (CL6 and CL7) showed largely overlapping genomic and transcriptome expression profiles when compared to the heterogeneous maternal culture and to e ach other. In accordance to the CSC hypothesis, both clones showed neural stem/progenitor traits upon analysis in the mcNSA, many of the expressed genes would be expected active in bona fide developing CNS neural stem cells, and the respective chromosomal aberrations were GBM typic. Yet, CL6 and CL7 cells differed in their ability to induce tumors in orthotopic xenograft experiments. The characteristic loss of KLF6 expression due to a distinctive copy neural LOH on chromosome 10p could explain this marked b iological behavior of CL6 cells. These findings attenuate previous implications on tumor suppressing activities of KLF625 to an individual cell derived level, and furthermore demonstrate the need to explore single cells and their progeny for classification of distinctive targets in malignant brain tumor diagnosis and therapy. An intriguing observation of our study, echoing conclusions made by Beier et al., 2007, was that the ability for long term propagation and the expression of CD133 two
149 candidate chara cteristics of tumor sustaining cells according to existing theories, do not represent essential hallmarks of tumorigenicity. Although both cell lines robust expressed CD133, neither heterogeneous 018 AE cultures, nor single cell derived 019 GBM CL7 cells displayed tumorigenic potential. These findings weaken the argument within recent discussions relating to the specificity of CD133 as an indicator of tumorigenic cells (Beier et al., 2007; Shmelkov et al., 2008). Alternatively this result may be viewed in light of the potential pitfalls of xenograft experimentation. Just as Kelly et al., 2007 suggested, there might be a need to co transfer additional accessory cells to enable complete mimicking of malignant disease. This thought is further driven by our pre liminary observations of tumorigenic CL6 and non tumorigenic CL7 cells relying on each other for prolonged self renewal. Cellular interactions might involve components of the pleiotropic Notch WNT pathways, for which partly complementary expre ssion profiles were recorded. And consequently, a potential lack of cellular interactions could explain why, in our experiments, CL6 cells required more time to elicit neuro behavioral abnormalities in xenografted animals than similar quantities of heterog eneous 019LPOmig cells (715 vs. 413 d, respectively. Lastly, the particular phenotypic constellation of CD133 /MGMTlow CL6 and CD133+/MGMT+ CL7 cells deserves attention, as MGMT expressing GBM cells are considered less susceptible to alkylating chemothe rapy. Taken together, our data exemplify an alternative in vitro approach suited to study the cellular basis of heterogeneity typically observed in malignant tumors, such as GBM3. While, currently, the entire spectrum of clonally active cellular phenotype s cannot be anticipated, their co existence needs to be considered, as well as their
150 interactions with each other, and the surrounding tissue. Our method may help to interrogate cancers for the particular stem cell types at the root of tumor formation an d tumor maintenance. Knowledge of the identity of these cells as well as a more complete understanding of their interactions with others cells within and without the tumor will help to direct our therapeutic efforts in targeting these unique cells. Respec tive analysis relating phenotypic, molecular, genetic and epigenetic traits to functional activity may pave the way for dissecting the hierarchical composition of human malignancies, facilitating their classification and leading to a more individualized di agnostic and therapeutic approach.
151 LIST OF REFERENCES Adams JM, Strasser A (2008) Is tumor growth sustained by rare cancer stem cells or dominant clones? Cancer Res 68:4018 4021. Altman J (1962) Are new neurons formed in the brains of adult mammals? Sci ence 135:1127 1128. Alvarez Buylla A, Lim DA (2004) For the long run: maintaining germinal niches in the adult brain. Neuron 41:683 686. Anderson MJ, Waxman SG (1985) Neurogenesis in adult vertebrate spinal cord in situ and in vitro: a new model system. An n N Y Acad Sci 457:213 233. Bachoo RM, Maher EA, Ligon KL, Sharpless NE, Chan SS, You MJ, Tang Y, DeFrances J, Stover E, Weissleder R, Rowitch DH, Louis DN, DePinho RA (2002) Epidermal growth factor receptor and Ink4a/Arf: convergent mechanisms governing t erminal differentiation and transformation along the neural stem cell to astrocyte axis. Cancer Cell 1:269 277. Bao S, Wu Q, McLendon RE, Hao Y, Shi Q, Hjelmeland AB, Dewhirst MW, Bigner DD, Rich JN (2006) Glioma stem cells promote radioresistance by prefe rential activation of the DNA damage response. Nature 444:756 760. Barres BA (2003) What is a glial cell? Glia 43:4 5. Beier D, Hau P, Proescholdt M, Lohmeier A, Wischhusen J, Oefner PJ, Aigner L, Brawanski A, Bogdahn U, Beier CP (2007) CD133(+) and CD133( ) glioblastoma derived cancer stem cells show differential growth characteristics and molecular profiles. Cancer Res 67:4010 4015. Brabletz S, Brabletz T (2010) The ZEB/miR 200 feedback loop -a motor of cellular plasticity in development and cancer? EMBO Rep 11:670 677. Brodkey JA, Laywell ED, O'Brien TF, Faissner A, Stefansson K, Dorries HU, Schachner M, Steindler DA (1995) Focal brain injury and upregulation of a developmentally regulated extracellular matrix protein. J Neurosurg 82:106 112. Bruni JE (19 98) Ependymal development, proliferation, and functions: a review. Microsc Res Tech 41:2 13. Buffo A, Rite I, Tripathi P, Lepier A, Colak D, Horn AP, Mori T, Gotz M (2008) Origin and progeny of reactive gliosis: A source of multipotent cells in the injured brain. Proc Natl Acad Sci U S A 105:3581 3586. Charles NA, Holland EC, Gilbertson R, Glass R, Kettenmann H (2011) The brain tumor microenvironment. Glia 59:1169 1180.
152 Chen Y, Miles DK, Hoang T, Shi J, Hurlock E, Kernie SG, Lu QR (2008) The basic helix loo p helix transcription factor olig2 is critical for reactive astrocyte proliferation after cortical injury. J Neurosci 28:10983 10989. Chiasson BJ, Tropepe V, Morshead CM, van der Kooy D (1999) Adult mammalian forebrain ependymal and subependymal cells demo nstrate proliferative potential, but only subependymal cells have neural stem cell characteristics. J Neurosci 19:4462 4471. Cooper NG, Steindler DA (1986a) Monoclonal antibody to glial fibrillary acidic protein reveals a parcellation of individual barrels in the early postnatal mouse somatosensory cortex. Brain Res 380:341 348. Cooper NG, Steindler DA (1986b) Lectins demarcate the barrel subfield in the somatosensory cortex of the early postnatal mouse. J Comp Neurol 249:157 169. Cooper NG, Steindler DA (1 989) Critical period dependent alterations of the transient body image in the rodent cerebral cortex. Brain Res 489:167 176. Danilov AI, Gomes Leal W, Ahlenius H, Kokaia Z, Carlemalm E, Lindvall O (2009) Ultrastructural and antigenic properties of neural s tem cells and their progeny in adult rat subventricular zone. Glia 57:136 152. DeAngelis LM (2001) Brain tumors. N Engl J Med 344:114 123. Deleyrolle LP, Harding A, Cato K, Siebzehnrubl FA, Rahman M, Azari H, Olson S, Gabrielli B, Osborne G, Vescovi A, Rey nolds BA (2011) Evidence for label retaining tumour initiating cells in human glioblastoma. Brain 134:1331 1343. Doetsch F, Garcia Verdugo JM, Alvarez Buylla A (1997) Cellular composition and three dimensional organization of the subventricular germinal zo ne in the adult mammalian brain. J Neurosci 17:5046 5061. Doetsch F, Garcia Verdugo JM, Alvarez Buylla A (1999a) Regeneration of a germinal layer in the adult mammalian brain. Proc Natl Acad Sci U S A 96:11619 11624. Doetsch F, Caille I, Lim DA, Garcia Ver dugo JM, Alvarez Buylla A (1999b) Subventricular zone astrocytes are neural stem cells in the adult mammalian brain. Cell 97:703 716. Edwards LA, Woolard K, Son MJ, Li A, Lee J, Ene C, Mantey SA, Maric D, Song H, Belova G, Jensen RT, Zhang W, Fine HA (2011 ) Effect of brain and tumor derived connective tissue growth factor on glioma invasion. J Natl Cancer Inst 103:1162 1178. Egar M, Singer M (1972) The role of ependyma in spinal cord regeneration in the urodele, Triturus. Exp Neurol 37:422 430.
153 Fisher D, X ing B, Dill J, Li H, Hoang HH, Zhao Z, Yang XL, Bachoo R, Cannon S, Longo FM, Sheng M, Silver J, Li S (2011) Leukocyte common antigen related phosphatase is a functional receptor for chondroitin sulfate proteoglycan axon growth inhibitors. J Neurosci 31:14 051 14066. Font E, Desfilis E, Perez Canellas M, Alcantara S, Garcia Verdugo JM (1997) 3 Acetylpyridine induced degeneration and regeneration in the adult lizard brain: a qualitative and quantitative analysis. Brain Res 754:245 259. Fraser MM, Zhu X, Kwon CH, Uhlmann EJ, Gutmann DH, Baker SJ (2004) Pten loss causes hypertrophy and increased proliferation of astrocytes in vivo. Cancer Res 64:7773 7779. Gaiano N, Nye JS, Fishell G (2000) Radial glial identity is promoted by Notch1 signaling in the murine fore brain. Neuron 26:395 404. Galli R, Binda E, Orfanelli U, Cipelletti B, Gritti A, De Vitis S, Fiocco R, Foroni C, Dimeco F, Vescovi A (2004) Isolation and characterization of tumorigenic, stem like neural precursors from human glioblastoma. Cancer Res 64:70 11 7021. Galtrey CM, Fawcett JW (2007) The role of chondroitin sulfate proteoglycans in regeneration and plasticity in the central nervous system. Brain Res Rev 54:1 18. Garcia AD, Doan NB, Imura T, Bush TG, Sofroniew MV (2004) GFAP expressing progenitors are the principal source of constitutive neurogenesis in adult mouse forebrain. Nat Neurosci 7:1233 1241. Gates MA, Thomas LB, Howard EM, Laywell ED, Sajin B, Faissner A, Gotz B, Silver J, Steindler DA (1995) Cell and molecular analysis of the developing a nd adult mouse subventricular zone of the cerebral hemispheres. J Comp Neurol 361:249 266. Glas M, Rath BH, Simon M, Reinartz R, Schramme A, Trageser D, Eisenreich R, Leinhaas A, Keller M, Schildhaus HU, Garbe S, Steinfarz B, Pietsch T, Steindler DA, Schra mm J, Herrlinger U, Brustle O, Scheffler B (2010) Residual tumor cells are unique cellular targets in glioblastoma. Ann Neurol 68:264 269. Gonzalez Mde L, Malemud CJ, Silver J (1993) Role of astroglial extracellular matrix in the formation of rat olfactory bulb glomeruli. Exp Neurol 123:91 105. Hambardzumyan D, Cheng YK, Haeno H, Holland EC, Michor F (2011) The probable cell of origin of NF1 and PDGF driven glioblastomas. PLoS One 6:e24454. Hartfuss E, Galli R, Heins N, Gotz M (2001) Characterization of CN S precursor subtypes and radial glia. Dev Biol 229:15 30. Hasegawa K, Grumet M (2003) Trauma induced tumorigenesis of cells implanted into the rat spinal cord. J Neurosurg 98:1065 1071.
154 Hemmati HD, Nakano I, Lazareff JA, Masterman Smith M, Geschwind DH, Br onner Fraser M, Kornblum HI (2003) Cancerous stem cells can arise from pediatric brain tumors. Proc Natl Acad Sci U S A 100:15178 15183. Holsken A, Eyupoglu IY, Lueders M, Trankle C, Dieckmann D, Buslei R, Hahnen E, Blumcke I, Siebzehnrubl FA (2006) Ex viv o therapy of malignant melanomas transplanted into organotypic brain slice cultures using inhibitors of histone deacetylases. Acta Neuropathol 112:205 215. Ignatova TN, Kukekov VG, Laywell ED, Suslov ON, Vrionis FD, Steindler DA (2002) Human cortical glial tumors contain neural stem like cells expressing astroglial and neuronal markers in vitro. Glia 39:193 206. Imura T, Kornblum HI, Sofroniew MV (2003) The predominant neural stem cell isolated from postnatal and adult forebrain but not early embryonic fore brain expresses GFAP. J Neurosci 23:2824 2832. Jin Y, Ketschek A, Jiang Z, Smith G, Fischer I (2011) Chondroitinase activity can be transduced by a lentiviral vector in vitro and in vivo. J Neurosci Methods 199:208 213. Johansson CB, Momma S, Clarke DL, Ri sling M, Lendahl U, Frisen J (1999) Identification of a neural stem cell in the adult mammalian central nervous system. Cell 96:25 34. Kempermann G, Kuhn HG, Gage FH (1997) More hippocampal neurons in adult mice living in an enriched environment. Nature 38 6:493 495. Kerever A, Schnack J, Vellinga D, Ichikawa N, Moon C, Arikawa Hirasawa E, Efird JT, Mercier F (2007) Novel extracellular matrix structures in the neural stem cell niche capture the neurogenic factor fibroblast growth factor 2 from the extracellu lar milieu. Stem Cells 25:2146 2157. Kimmelman AC, Qiao RF, Narla G, Banno A, Lau N, Bos PD, Nunez Rodriguez N, Liang BC, Guha A, Martignetti JA, Friedman SL, Chan AM (2004) Suppression of glioblastoma tumorigenicity by the Kruppel like transcription facto r KLF6. Oncogene 23:5077 5083. Kishi K (1987) Golgi studies on the development of granule cells of the rat olfactory bulb with reference to migration in the subependymal layer. J Comp Neurol 258:112 124. Kitada M, Chakrabortty S, Matsumoto N, Taketomi M, I de C (2001) Differentiation of choroid plexus ependymal cells into astrocytes after grafting into the pre lesioned spinal cord in mice. Glia 36:364 374. Kondo T, Raff M (2000) Oligodendrocyte precursor cells reprogrammed to become multipotential CNS stem c ells. Science 289:1754 1757.
155 Laywell ED, Steindler DA (1991) Boundaries and wounds, glia and glycoconjugates. Cellular and molecular analyses of developmental partitions and adult brain lesions. Ann N Y Acad Sci 633:122 141. Laywell ED, Rakic P, Kukekov VG Holland EC, Steindler DA (2000) Identification of a multipotent astrocytic stem cell in the immature and adult mouse brain. Proc Natl Acad Sci U S A 97:13883 13888. Laywell ED, Dorries U, Bartsch U, Faissner A, Schachner M, Steindler DA (1992) Enhanced e xpression of the developmentally regulated extracellular matrix molecule tenascin following adult brain injury. Proc Natl Acad Sci U S A 89:2634 2638. Laywell ED, Kearns SM, Zheng T, Chen KA, Deng J, Chen HX, Roper SN, Steindler DA (2005) Neuron to astrocy te transition: phenotypic fluidity and the formation of hybrid asterons in differentiating neurospheres. J Comp Neurol 493:321 333. Lee H, McKeon RJ, Bellamkonda RV (2010) Sustained delivery of thermostabilized chABC enhances axonal sprouting and functiona l recovery after spinal cord injury. Proc Natl Acad Sci U S A 107:3340 3345. Levitt P, Rakic P (1980) Immunoperoxidase localization of glial fibrillary acidic protein in radial glial cells and astrocytes of the developing rhesus monkey brain. J Comp Neurol 193:815 840. Ligon KL, Huillard E, Mehta S, Kesari S, Liu H, Alberta JA, Bachoo RM, Kane M, Louis DN, Depinho RA, Anderson DJ, Stiles CD, Rowitch DH (2007) Olig2 regulated lineage restricted pathway controls replication competence in neural stem cells and malignant glioma. Neuron 53:503 517. Lin RC, Matesic DF, Marvin M, McKay RD, Brustle O (1995) Re expression of the intermediate filament nestin in reactive astrocytes. Neurobiol Dis 2:79 85. Louis DN, Ohgaki H, Wiestler OD, Cavenee WK, Burger PC, Jouvet A Scheithauer BW, Kleihues P (2007) The 2007 WHO classification of tumours of the central nervous system. Acta Neuropathol 114:97 109. Luskin MB (1993) Restricted proliferation and migration of postnatally generated neurons derived from the forebrain subve ntricular zone. Neuron 11:173 189. Markovic DS, Vinnakota K, Chirasani S, Synowitz M, Raguet H, Stock K, Sliwa M, Lehmann S, Kalin R, van Rooijen N, Holmbeck K, Heppner FL, Kiwit J, Matyash V, Lehnardt S, Kaminska B, Glass R, Kettenmann H (2009) Gliomas in duce and exploit microglial MT1 MMP expression for tumor expansion. Proc Natl Acad Sci U S A 106:12530 12535.
156 Marshall GP, 2nd, Demir M, Steindler DA, Laywell ED (2008) Subventricular zone microglia possess a unique capacity for massive in vitro expansion. Glia 56:1799 1808. Mazlo M, Gasz B, Szigeti A, Zsombok A, Gallyas F (2004) Debris of "dark" (compacted) neurones are removed from an otherwise undamaged environment mainly by astrocytes via blood vessels. J Neurocytol 33:557 567. McKeon RJ, Schreiber RC, Rudge JS, Silver J (1991) Reduction of neurite outgrowth in a model of glial scarring following CNS injury is correlated with the expression of inhibitory molecules on reactive astrocytes. J Neurosci 11:3398 3411. Meletis K, Barnabe Heider F, Carlen M, Eve rgren E, Tomilin N, Shupliakov O, Frisen J (2008) Spinal cord injury reveals multilineage differentiation of ependymal cells. PLoS Biol 6:e182. Mercier F, Kitasako JT, Hatton GI (2002) Anatomy of the brain neurogenic zones revisited: fractones and the fibr oblast/macrophage network. J Comp Neurol 451:170 188. Merkle FT, Mirzadeh Z, Alvarez Buylla A (2007) Mosaic organization of neural stem cells in the adult brain. Science 317:381 384. Merkle FT, Tramontin AD, Garcia Verdugo JM, Alvarez Buylla A (2004) Radia l glia give rise to adult neural stem cells in the subventricular zone. Proc Natl Acad Sci U S A 101:17528 17532. Miskevich F (1999) Laminar redistribution of a glial subtype in the chick optic tectum. Brain Res Dev Brain Res 115:103 109. Misson JP, Austin CP, Takahashi T, Cepko CL, Caviness VS, Jr. (1991) The alignment of migrating neural cells in relation to the murine neopallial radial glial fiber system. Cereb Cortex 1:221 229. Molowny A, Nacher J, Lopez Garcia C (1995) Reactive neurogenesis during rege neration of the lesioned medial cerebral cortex of lizards. Neuroscience 68:823 836. Morshead CM, van der Kooy D (1992) Postmitotic death is the fate of constitutively proliferating cells in the subependymal layer of the adult mouse brain. J Neurosci 12:24 9 256. Nelson CM, Bissell MJ (2006) Of extracellular matrix, scaffolds, and signaling: tissue architecture regulates development, homeostasis, and cancer. Annu Rev Cell Dev Biol 22:287 309.
157 Noctor SC, Flint AC, Weissman TA, Dammerman RS, Kriegstein AR (200 1) Neurons derived from radial glial cells establish radial units in neocortex. Nature 409:714 720. Noctor SC, Flint AC, Weissman TA, Wong WS, Clinton BK, Kriegstein AR (2002) Dividing precursor cells of the embryonic cortical ventricular zone have morphol ogical and molecular characteristics of radial glia. J Neurosci 22:3161 3173. Nordlander RH, Singer M (1978) The role of ependyma in regeneration of the spinal cord in the urodele amphibian tail. J Comp Neurol 180:349 374. Nunes MC, Roy NS, Keyoung HM, Goo dman RR, McKhann G, 2nd, Jiang L, Kang J, Nedergaard M, Goldman SA (2003) Identification and isolation of multipotential neural progenitor cells from the subcortical white matter of the adult human brain. Nat Med 9:439 447. Okano Uchida T, Himi T, Komiya Y Ishizaki Y (2004) Cerebellar granule cell precursors can differentiate into astroglial cells. Proc Natl Acad Sci U S A 101:1211 1216. Osawa M, Hanada K, Hamada H, Nakauchi H (1996) Long term lymphohematopoietic reconstitution by a single CD34 low/negativ e hematopoietic stem cell. Science 273:242 245. Pardal R, Clarke MF, Morrison SJ (2003) Applying the principles of stem cell biology to cancer. Nat Rev Cancer 3:895 902. Passegue E, Jamieson CH, Ailles LE, Weissman IL (2003) Normal and leukemic hematopoies is: are leukemias a stem cell disorder or a reacquisition of stem cell characteristics? Proc Natl Acad Sci U S A 100 Suppl 1:11842 11849. Perez Canellas MM, Garcia Verdugo JM (1996) Adult neurogenesis in the telencephalon of a lizard: a [3H]thymidine autor adiographic and bromodeoxyuridine immunocytochemical study. Brain Res Dev Brain Res 93:49 61. Petreanu L, Alvarez Buylla A (2002) Maturation and death of adult born olfactory bulb granule neurons: role of olfaction. J Neurosci 22:6106 6113. Piccirillo SG, Reynolds BA, Zanetti N, Lamorte G, Binda E, Broggi G, Brem H, Olivi A, Dimeco F, Vescovi AL (2006) Bone morphogenetic proteins inhibit the tumorigenic potential of human brain tumour initiating cells. Nature 444:761 765. Ponten J, Macintyre EH (1968) Long term culture of normal and neoplastic human glia. Acta Pathol Microbiol Scand 74:465 486. Potten CS, Loeffler M (1990) Stem cells: attributes, cycles, spirals, pitfalls and uncertainties. Lessons for and from the crypt. Development 110:1001 1020.
158 Quinones Hinojosa A, Chaichana K (2007) The human subventricular zone: a source of new cells and a potential source of brain tumors. Exp Neurol 205:313 324. Quinones Hinojosa A, Sanai N, Soriano Navarro M, Gonzalez Perez O, Mirzadeh Z, Gil Perotin S, Romero Rodrigu ez R, Berger MS, Garcia Verdugo JM, Alvarez Buylla A (2006) Cellular composition and cytoarchitecture of the adult human subventricular zone: a niche of neural stem cells. J Comp Neurol 494:415 434. Ramn y Cajal S, May RM (1928) Degeneration & regeneratio n of the nervous system. London,: Oxford university press, Humphrey Milford. Rao JS (2003) Molecular mechanisms of glioma invasiveness: the role of proteases. Nat Rev Cancer 3:489 501. Reardon DA, Zolutsky MR, Bigner DD (2007) Antitenascin C monoclonal ant ibody radioimmunotherapy for malignant glioma patients. Expert Rev Anticanc 7:675 687. Reier PJ, Houle JD (1988) The glial scar: its bearing on axonal elongation and transplantation approaches to CNS repair. Adv Neurol 47:87 138. Reynolds BA, Rietze RL (20 05) Neural stem cells and neurospheres -re evaluating the relationship. Nat Methods 2:333 336. Rietze RL, Reynolds BA (2006) Neural stem cell isolation and characterization. Methods Enzymol 419:3 23. Sakakibara S, Okano H (1997) Expression of neural RNA bi nding proteins in the postnatal CNS: implications of their roles in neuronal and glial cell development. J Neurosci 17:8300 8312. Scheffler B, Walton NM, Lin DD, Goetz AK, Enikolopov G, Roper SN, Steindler DA (2005) Phenotypic and functional characterizati on of adult brain neuropoiesis. Proc Natl Acad Sci U S A 102:9353 9358. Scherer HJ (1940) A Critical Review: The Pathology of Cerebral Gliomas. J Neurol Psychiatry 3:147 177. Seri B, Garcia Verdugo JM, McEwen BS, Alvarez Buylla A (2001) Astrocytes give ris e to new neurons in the adult mammalian hippocampus. J Neurosci 21:7153 7160. Shmelkov SV, Butler JM, Hooper AT, Hormigo A, Kushner J, Milde T, St Clair R, Baljevic M, White I, Jin DK, Chadburn A, Murphy AJ, Valenzuela DM, Gale NW, Thurston G, Yancopoulos GD, D'Angelica M, Kemeny N, Lyden D, Rafii S (2008) CD133 expression is not restricted to stem cells, and both CD133+ and CD133 metastatic colon cancer cells initiate tumors. J Clin Invest 118:2111 2120.
159 Siebzehnrubl FA, Blumcke I (2008) Neurogenesis in t he human hippocampus and its relevance to temporal lobe epilepsies. Epilepsia 49 Suppl 5:55 65. Siebzehnrubl FA, Jeske I, Muller D, Buslei R, Coras R, Hahnen E, Huttner HB, Corbeil D, Kaesbauer J, Appl T, von Horsten S, Blumcke I (2009) Spontaneous in vitr o transformation of adult neural precursors into stem like cancer cells. Brain Pathol 19:399 408. Silver DJ, Steindler DA (2009) Common astrocytic programs during brain development, injury and cancer. Trends Neurosci 32:303 311. Silver J, Miller JH (2004) Regeneration beyond the glial scar. Nat Rev Neurosci 5:146 156. Simpson SB, Jr., Duffy MT (1994) The lizard spinal cord: a model system for the study of spinal cord injury and repair. Prog Brain Res 103:229 241. Singh SK, Hawkins C, Clarke ID, Squire JA, B ayani J, Hide T, Henkelman RM, Cusimano MD, Dirks PB (2004) Identification of human brain tumour initiating cells. Nature 432:396 401. Snow DM, Steindler DA, Silver J (1990) Molecular and cellular characterization of the glial roof plate of the spinal cord and optic tectum: a possible role for a proteoglycan in the development of an axon barrier. Dev Biol 138:359 376. Spassky N, Merkle FT, Flames N, Tramontin AD, Garcia Verdugo JM, Alvarez Buylla A (2005) Adult ependymal cells are postmitotic and are derive d from radial glial cells during embryogenesis. J Neurosci 25:10 18. Steindler DA (1993) Glial boundaries in the developing nervous system. Annu Rev Neurosci 16:445 470. Steindler DA, Laywell ED (2003) Astrocytes as stem cells: nomenclature, phenotype, and translation. Glia 43:62 69. Steindler DA, O'Brien TF, Cooper NG (1988) Glycoconjugate boundaries during early postnatal development of the neostriatal mosaic. J Comp Neurol 267:357 369. Steindler DA, Kukekov VG, Thomas LB, Fillmore H, Suslov O, Scheffler B, O'Brien TF, Kusakabe M, Laywell ED (1998) Boundary molecules during brain development, injury, and persistent neurogenesis -in vivo and in vitro studies. Prog Brain Res 117:179 196. Suslov ON, Kukekov VG, Ignatova TN, Steindler DA (2002) Neural stem cel l heterogeneity demonstrated by molecular phenotyping of clonal neurospheres. Proc Natl Acad Sci U S A 99:14506 14511.
160 Takahashi K, Tanabe K, Ohnuki M, Narita M, Ichisaka T, Tomoda K, Yamanaka S (2007) Induction of pluripotent stem cells from adult human f ibroblasts by defined factors. Cell 131:861 872. Takahashi M, Arai Y, Kurosawa H, Sueyoshi N, Shirai S (2003) Ependymal cell reactions in spinal cord segments after compression injury in adult rat. J Neuropathol Exp Neurol 62:185 194. Taupin P, Gage FH (20 02) Adult neurogenesis and neural stem cells of the central nervous system in mammals. J Neurosci Res 69:745 749. Tester NJ, Plaas AH, Howland DR (2007) Effect of body temperature on chondroitinase ABC's ability to cleave chondroitin sulfate glycosaminogly cans. J Neurosci Res 85:1110 1118. Thomas LB, Gates MA, Steindler DA (1996) Young neurons from the adult subependymal zone proliferate and migrate along an astrocyte, extracellular matrix rich pathway. Glia 17:1 14. Tom VJ, Steinmetz MP, Miller JH, Doller CM, Silver J (2004) Studies on the development and behavior of the dystrophic growth cone, the hallmark of regeneration failure, in an in vitro model of the glial scar and after spinal cord injury. J Neurosci 24:6531 6539. Uhrbom L, Dai C, Celestino JC, Ro senblum MK, Fuller GN, Holland EC (2002) Ink4a Arf loss cooperates with KRas activation in astrocytes and neural progenitors to generate glioblastomas of various morphologies depending on activated Akt. Cancer Res 62:5551 5558. van Praag H, Christie BR, Se jnowski TJ, Gage FH (1999) Running enhances neurogenesis, learning, and long term potentiation in mice. Proc Natl Acad Sci U S A 96:13427 13431. van Praag H, Schinder AF, Christie BR, Toni N, Palmer TD, Gage FH (2002) Functional neurogenesis in the adult h ippocampus. Nature 415:1030 1034. Vescovi AL, Galli R, Reynolds BA (2006) Brain tumour stem cells. Nat Rev Cancer 6:425 436. Viapiano MS, Bi WL, Piepmeier J, Hockfield S, Matthews RT (2005) Novel tumor specific isoforms of BEHAB/brevican identified in huma n malignant gliomas. Cancer Res 65:6726 6733. Weiss S, Dunne C, Hewson J, Wohl C, Wheatley M, Peterson AC, Reynolds BA (1996) Multipotent CNS stem cells are present in the adult mammalian spinal cord and ventricular neuroaxis. J Neurosci 16:7599 7609.
161 Well ner U, Schubert J, Burk UC, Schmalhofer O, Zhu F, Sonntag A, Waldvogel B, Vannier C, Darling D, zur Hausen A, Brunton VG, Morton J, Sansom O, Schuler J, Stemmler MP, Herzberger C, Hopt U, Keck T, Brabletz S, Brabletz T (2009) The EMT activator ZEB1 promote s tumorigenicity by repressing stemness inhibiting microRNAs. Nat Cell Biol 11:1487 1495. Wolf K, Wu YI, Liu Y, Geiger J, Tam E, Overall C, Stack MS, Friedl P (2007) Multi step pericellular proteolysis controls the transition from individual to collective cancer cell invasion. Nat Cell Biol 9:893 904. Yu J, Vodyanik MA, Smuga Otto K, Antosiewicz Bourget J, Frane JL, Tian S, Nie J, Jonsdottir GA, Ruotti V, Stewart R, Slukvin, II, Thomson JA (2007) Induced pluripotent stem cell lines derived from human somati c cells. Science 318:1917 1920. Zhang H, Kelly G, Zerillo C, Jaworski DM, Hockfield S (1998) Expression of a cleaved brain specific extracellular matrix protein mediates glioma cell invasion In vivo. J Neurosci 18:2370 2376. Zheng T, Steindler DA, Laywell ED (2002) Transplantation of an indigenous neural stem cell population leading to hyperplasia and atypical integration. Cloning Stem Cells 4:3 8. Zheng T, Marshall GP, 2nd, Laywell ED, Steindler DA (2006) Neurogenic astrocytes transplanted into the adult m ouse lateral ventricle contribute to olfactory neurogenesis, and reveal a novel intrinsic subependymal neuron. Neuroscience 142:175 185.
162 BIOGRAPHICAL SKETCH Daniel was born in Boston, Massachusetts, raised in Cleveland, Ohio and currently lives in Gaines ville, Florida with his wife Melinda and daughter Olivia. He attended Charles F. Brush High School in South Euclid, Ohio and graduated magna cum laude with a BS in chemistry from Cleveland State University. Daniel worked for several years after graduati on as an organic synthesis chemist for the Cleveland biotech company Athersys, Inc. After leaving Athersys, Daniel went to work as a lab technician for his father Jerry, at Case Western Reserve University were he published his first papers and fell in lov e with biology. While working at CWRU, Daniel met his mentor Dennis Steindler at a Society for Neuroscience meeting in San Diego, California in 2004. The following fall, Daniel began his doctoral training at the University of Florida with Dennis as his advisor. This dissertation represents the culmination of his time as a student with Dennis. He and his family will leave Gainesville for New York City, where he has accepted oan Kettering Cancer Center.