University of Groningen. New neurons in the adult brain van der Borght, Karin

Transcription

1 University of Groningen New neurons in the adult brain van der Borght, Karin IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below. Document Version Publisher's PDF, also known as Version of record Publication date: 2006 Link to publication in University of Groningen/UMCG research database Citation for published version (APA): van der Borght, K. (2006). New neurons in the adult brain: A study on the regulation and function of neurogenesis in the adult rodent hippocampus. s.n. Copyright Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons). Take-down policy If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim. Downloaded from the University of Groningen/UMCG research database (Pure): For technical reasons the number of authors shown on this cover page is limited to 10 maximum. Download date:

2 New neurons in the adult brain A study on the regulation and function of neurogenesis in the adult rodent hippocampus Karin van der Borght

3

4 RIJKSUNIVERSITEIT GRONINGEN New neurons in the adult brain A study on the regulation and function of neurogenesis in the adult rodent hippocampus Proefschrift ter verkrijging van het doctoraat in de Wiskunde en Natuurwetenschappen aan de Rijksuniversiteit Groningen op gezag van de Rector Magnificus, dr. F. Zwarts, in het openbaar te verdedigen op vrijdag 13 januari 2006 om uur door Karin van der Borght geboren op 3 december 1978 te Boxtel

5 Promotor: Co-promotores: Prof. dr. P.G.M. Luiten Dr. E.A. van der Zee Dr. B.J.L. Eggen Beoordelingscommissie: Prof. dr. H.W. Boddeke Prof. dr. J.A. den Boer Prof. dr. J.M. Koolhaas ISBN x

6 The truth is rarely pure and never simple Oscar Wilde, The Importance of Being Earnest, 1895

7 The studies described in this thesis were carried out at the Department of Molecular Neurobiology, University of Groningen, The Netherlands and were financially supported by the Graduate school of Behavioral and Cognitive Neurosciences (BCN), University of Groningen. The printing of this thesis was fi nancially supported by Alzheimer Nederland. Cover: Doublecortin-positive cells in the dentate gyrus. Design: Jeroen en Karin van der Borght ( Printed by: Grafi sche bedrijf Ponsen & Looijen, Wageningen, The Netherlands

8 Contents Chapter 1: General Introduction 9 Chapter 2: Hippocampal cell proliferation across the day: Increase 29 by running wheel activity, but no effect of sleep and wakefulness Chapter 3: Differential regulation of hippocampal cell proliferation 41 and cell survival by physical exercise Chapter 4: Input from the medial septum regulates adult 51 hippocampal neurogenesis Chapter 5: Effects of active shock avoidance learning on hippocampal 63 neurogenesis and plasma levels of corticosterone Chapter 6: Morris water maze learning in two rat strains increases 77 PSA-NCAM expression in the dentate gyrus, but has no effect on hippocampal neurogenesis Chapter 7: Memory retrieval reduces the number of newly formed 89 hippocampal neurons in mice with baseline and exercise-enhanced levels of neurogenesis Chapter 8: General Discussion 107 List of Abbreviations 137 Color Figures 138 References 143 Nederlandse Samenvatting 177 Bedankt! 181 Curriculum Vitae 184 List of Publications 185

9 8

10 Chapter 1 General Introduction 9

11 Chapter 1 CONTENTS 1 History of neurogenesis research 2 The dentate gyrus 2.1 Anatomy 2.2 Input pathways to the DG 2.3 The septohippocampal pathway 3 Hippocampal neurogenesis: from stem cell to functional granule neuron 3.1 The putative hippocampal stem cell 3.2 Different subsets of proliferating precursor cells 3.3 Maturation of the newly formed cells 3.4 Fluctuations in newly formed cell number 4 Regulation of hippocampal neurogenesis 4.1 The neurogenic microenvironment 4.2 Internal and external factors that regulate hippocampal neurogenesis 4.3 Effects of physical exercise on hippocampal neurogenesis 5 Neurogenesis and learning 5.1 Environmental impact on hippocampal neurogenesis 5.2 Effects of learning on neurogenesis 5.3 Effects of reduced neurogenesis on learning 5.4 Computational models of neurogenesis: neurogenesis and memory clearance 6 Outline of this thesis 10

12 General introduction 1. History of neurogenesis research For over 100 years, the generally accepted view has been that the adult brain only contains postmitotic neurons and is incapable of generating new ones (Gross, 2000; Ramon y Cajal, 1928). The introduction of [3H]-thymidine, which is incorporated in the DNA of mitotic cells during the S-phase of the cell cycle, resulted in the discovery of newly formed cells in the adult rat brain (Altman and Das, 1965). These fi ndings have been largely neglected for many years, mainly because the available techniques at that time could not indisputably demonstrate that the labeled cells were indeed new neurons. In the 1980s, electron microscopical studies demonstrated that newly formed cells in the dentate gyrus (DG) and the olfactory bulb had the ultrastructural characteristics of new neurons (Kaplan and Hinds, 1977). A few years later, the presence of neurogenesis in the vocal control nucleus of canaries was discovered (Goldman and Nottebohm, 1983). Interest in the neurogenesis field started to increase dramatically when studies in male canaries demonstrated that cell death and cell recruitment in the vocal nuclei are seasonally regulated and play a role in song modification (Kirn et al., 1994). The introduction of 5-bromo-2 -deoxyuridine (BrdU) as a tool to study the generation of new cells caused a rapid expansion of the fi eld. BrdU is a thymidine analogue, which is incorporated in replicating DNA, and in contrast to [3H]-thymidine, BrdU can be visualized with immunocytochemistry. In addition, the use of BrdU as a tool to label dividing cells makes it possible to investigate the phenotype of newly formed cells with double labeling techniques (Gross, 2000). With these new techniques, adult hippocampal neurogenesis could be demonstrated in many species, including primates (Gould et al., 1999b; Kornack and Rakic, 1999) and even humans (Eriksson et al., 1998). Adult neurogenesis has been identified in the granule cell layer (GCL) of the hippocampus and in the olfactory bulb. New olfactory bulb neurons originate from undifferentiated progenitors that are located along the wall of the lateral ventricles, the subventricular zone (SVZ). Upon migration towards the olfactory bulb, via the so-called rostral migratory stream (RMS), they differentiate into interneurons (Bedard and Parent, 2004; Winner et al., 2002). In the hippocampus, an undifferentiated, proliferating cell population is located in the small layer between the GCL and the hilus, the subgranular zone (SGZ). After a few cell divisions, the proliferating cells become postmitotic, migrate into the granule cell layer, differentiate into neurons and become integrated into the existing hippocampal circuitry. The focus of this thesis is on hippocampal neurogenesis. Therefore, whenever the term adult neurogenesis is mentioned, this refers to neurogenesis in the adult DG. 11

13 Chapter 1 2. The dentate gyrus 2.1. Anatomy The dentate gyrus (DG) consists of a granule cell layer, containing densely packed granule cells (circa 1,000,000 in rats (Boss et al., 1985) and 250, ,000 granule cells in mice (Kempermann et al., 1997; Sun et al., 2004)), a molecular layer, in which the granule cell dendrites are located and a polymorphic layer or hilus that contains the axons of the granule cells, various types of interneurons and astrocytes. The bundle of axons from the granule cells that projects to the CA3 region is called the mossy fi ber pathway. Mossy fiber axons innervate two different cell types in the CA3 region, the pyramidal neurons (around 300,000 in rats (Amaral et al., 1990)) and interneurons, via morphologically distinct presynaptic terminals. About 50 granule neurons project to one CA3 pyramidal cell and around 14 pyramidal cells are innervated by one granule neuron via a specialized giant mossy fi ber bouton (Amaral and Dent, 1981). Each granule cell also innervates 40 to 50 CA3 interneurons via fi lopodial extensions that extend from the giant bouton and en passant boutons along the main mossy fi ber axon (Acsady et al., 1998; Frotscher et al., 1994). Therefore, granule cells actually innervate more interneurons than principal CA3 cells, suggesting that GABAergic cells are the major postsynaptic target of granule cells. However, although not many pyramidal neurons are activated by a single granule neuron, the connection appears to be strong enough to function as a detonator or teacher synapse for the storage of information in the CA3 network (Henze et al., 2002; Jung and McNaughton, 1993; Lisman, 1999). Each mossy fi ber axon also makes approximately synapses with cells of the hilus. The hilar cells are primarily inhibitory interneurons, but the mossy fi ber collaterals of a single granule cell make approximately 10 synapses with excitatory hilar mossy cells that project back to the granule cells forming a dense plexus in the inner molecular layer of the dentate gyrus (Amaral, 1978; Buckmaster et al., 1996; Henze et al., 2000). The GCL can be subdivided into the suprapyramidal (inner) blade and the infrapyramidal (outer) blade. During development of the DG, which takes place postnatally in rodents, cells of the infrapyramidal blade are generated somewhat later than those of the suprapyramidal layer (Schlessinger et al., 1975). Morphology of granule cells differs between both blades (Claiborne et al., 1990), with cells in the suprapyramidal blade having more total dendritic length, more dendritic segments and a greater transverse spread. Spine density of dendritic spines is greater on suprapyramidal cells than on infrapyramidal cells (Desmond and Levy, 1985). Combined with their greater dendritic length, these numbers indicate that suprapyramidal cells receive a signifi cantly greater number of terminals. On the other hand, stimulation of the molecular layer causes a larger population spike in the infrapyramidal blade than in the suprapyramidal blade and the suprapyramidal blade requires higher stimulus intensities to evoke EPSPs that reach the threshold for action potential generation (Scharfman et al., 2002) 12

14 General introduction 2.2. Input pathways to the DG The DG receives input from a variety of brain regions, a few of which are shortly being discussed here. The major cortical innervation of the hippocampus comes from the entorhinal cortex (EC). The axons of the cells that are located in layer II of the EC project axons to the outer two thirds of the molecular layer. These projections form the so-called perforant path. The EC-DG projections are not homogenously distributed over the septo-temporal pole of the hippocampus. Cell from the lateral EC innervate the septal part, whereas cells in the medial EC connect to the temporal pole of the dentate gyrus (Dolorfo and Amaral, 1998; Hjorth-Simonsen and Jeune, 1972). The inner third of the molecular layer is occupied by associational and commissural fi bers, which are intrinsic fi ber systems of the hippocampal formation (Amaral and Witter, 1989; Deller, 1998). The DG also receives dense monoaminergic input. Serotonergic neurons in the medial raphe innervate the hippocampus via the supracallosal pathway through the fi m- bria-fornix and via the infracallosal pathway through the cingulate bundle. Axons from the dorsal raphe nucleus reach the hippocampus via the supracallosal pathway or through the amygdala and the EC (Azmitia and Segal, 1978; Gulyas et al., 1999; Kohler et al., 1982). In addition, the DG shows strong immunoreactivity for tyrosine hydroxylase (TH), the rate-limiting enzyme for noradrenaline and dopamine synthesis. However, the source of TH that is present in the DG predominantly appears to originate from noradrenergic neurons in the locus coeruleus (Ishida et al., 2000; Moore and Bloom, 1979), and not from dopaminergic neurons (Gasbarri et al., 1994; Torack and Miller, 1994) The septohippocampal pathway One of the major afferent connections to the hippocampus is formed by the medial septum (MS). The fi bers that originate in the MS enter the hippocampus via the fi mbria-fornix and form a diffuse network in all hippocampal layers. Whereas the perforant path is thought to relay specific sensory cortical information about the environment to the hippocampus, the input from the MS is considered to be involved in providing information about the motivational, emotional and autonomic state of the animal (Gulyas et al., 1999). The MS consists of three cell types: cholinergic neurons, GABAergic projection neurons and GABAergic interneurons. The cholinergic neurons in the MS innervate all three cell types in the hippocampus: the pyramidal cells, granule neurons and interneurons (Frotscher, 1989; Gaykema et al., 1990). The GABAergic projection neurons also send axons to the hippocampal formation, where they make contact with interneurons (Freund, 1992; Frotscher et al., 1992; Gulyas et al., 1999). Electrophysiological studies indicate that the cholinergic input to the dentate granule neurons causes a direct excitation of these cells (Hortnagl and Hellweg, 1997; Wheal and Miller, 1980). The cholinergic projection to hippocampal interneurons, leads to an increased interneuronal activity (Frazier et al., 2003; Pitler and Alger, 1992). Also within the MS, there is a complex communication between the different cell types. Not only do the interneurons inhibit the cholinergic neurons, the cholinergic 13

15 Chapter 1 neurons also exert an effect on the GABAergic neurons, which have been shown to express muscarinic receptors (Brauer et al., 1998; Van der Zee and Luiten, 1994). The cholinergic cells in the MS also express muscarinic receptors, but in a lower density than GABAergic neurons (Van der Zee and Luiten, 1994). A subpopulation of cholinergic neurons is inhibited by muscarinic stimulation, suggesting the presence of the inhibitory m2 receptor subtype (Wu et al., 2000). Finally, the hippocampus also has reciprocal connections with the MS. These projections come either from hippocampal pyramidal cells that innervate cholinergic neurons in the MS via the lateral septum (Leranth and Frotscher, 1989; Risold and Swanson, 1997), or from GABAergic, primarily somatostatin-positive neurons (Jinno and Kosaka, 2002) that directly project to GABAergic neurons in the medial septum and, to a smaller extent, also to cholinergic cells (Gaykema et al., 1991; Schwerdtfeger and Buhl, 1986; Toth et al., 1993) 3. Hippocampal neurogenesis: from stem cell to functional granule neuron 3.1. The putative hippocampal stem cell Neurogenesis is thought to depend on the presence of neural stem cells (NSC s). Stem cells can be defi ned as undifferentiated cells that show asymmetric cell division, that have the capacity to self-renew and that are multipotent. Since no specific marker for NSC s has been identified yet, an NSC is usually defi ned as a cell that has the capacity to form multipotent neurospheres in vitro. Based on this criterion, it has been questioned whether the adult hippocampal formation does in fact contain NSC s, since there are reports showing that clonogenic spheres are only formed in cultures of the SVZ and not of the hippocampus (Seaberg and van der Kooy, 2002). Others, however, were able to obtain multipotent neurospheres from the adult hippocampus (Gage et al., 1995; Palmer et al., 1999). Moreover, the fact that new neurons and glia cells continue to be produced throughout an animal s life suggests the presence of NSC s in the hippocampus (Alvarez-Buylla et al., 2002). Detailed analysis of the subventricular germinal zone demonstrated that the putative stem cell in this region is likely to be GFAP-positive (Alvarez-Buylla et al., 2002; Seri et al., 2001). There is also increasing evidence that the hippocampal stem cell is a GFAPexpressing cell (Garcia et al., 2004; Kempermann et al., 2004; Seri et al., 2001; Seri et al., 2004). In the SGZ two different types of GFAP-positive cells can be found, horizontal and radial astrocytes (Seri et al., 2001; Seri et al., 2004). The radial astrocytes or type-1 cells (Fukuda et al., 2003; Kempermann et al., 2004) are, in contrast to the multipolar appearance of horizontal astrocytes, predominantly uni-or bipolar and they have a radial process that goes through the granule cell layer and short tangential processes extending along the SGZ (Alvarez-Buylla et al., 2002; Fukuda et al., 2003; Garcia et al., 2004). Also at the electronmicroscopical level, differences can be observed between radial and 14

16 General introduction horizontal astrocytes. Radial astrocytes have lighter mitochondria, more organelles, more polyribosomes and more prominent intermediate fi lament bundles in their processes (Seri et al., 2001). Next to GFAP, the radial astrocytes express vimentin (Garcia et al., 2004) and nestin (Fukuda et al., 2003; Kempermann et al., 2004; Mignone et al., 2004; Zhuo et al., 2001), which are not expressed by differentiated multipolar astrocytes. Radial astrocytes actually make up about two-thirds of the nestin-expressing population of cells in the SGZ. However, of the total nestin-expressing population, radial glial like cells are thought to form only 5% of the dividing cells (Kempermann et al., 2004). Evidence that radial astrocytes in the SGZ are indeed the precursor cells for granule neurons comes from studies in which SGZ astrocytes were retrovirally labeled. A few weeks after labeling, positive granule neurons in the GCL were observed, indicating that these new neurons were derived from GFAP-positive cells (Seri et al., 2001). Some astrocytes remained labeled, suggesting symmetric cell division into two astrocytes or asymmetric cell division, a key feature of stem cells (Alvarez-Buylla et al., 2002; Steiner et al., 2004). Additional evidence supporting the theory that NSC s in the hippocampus express GFAP comes from studies with a transgenic mouse model in which the proliferating population of GFAP-expressing cells in the forebrain could specifically and conditionally be deleted. Elimination of this cell population entirely prevented the formation of new hippocampal neurons (Garcia et al., 2004). Using GFAP-Cre reporter mice, the fate of GFAP-progenitors was identifi ed. Two weeks after injection with BrdU, over 90% of NeuN-expressing BrdU-positive cells expressed the reporter protein, indicating that the majority of the newly formed neurons in the DG are derived from GFAP-expressing progenitors Different subsets of precursor cells Radial astrocytes do not directly differentiate into mature granule neurons, but fi rst give rise to different subtypes of intermediate precursors. Based on their morphology (Fukuda et al., 2003; Seri et al., 2001; Seri et al., 2004), their marker expression (Fukuda et al., 2003; Jessberger and Kempermann, 2003; Kempermann et al., 2004; Kronenberg et al., 2003; Steiner et al., 2004) and their electrophysiological properties (Ambrogini et al., 2004a; Fukuda et al., 2003) various subtypes of precursors have been described, thereby using different nomenclature. A summary of the different precursor subtypes that can be found in the adult hippocampus is presented in Figure 1 (page 138). Astrocytic stem cells give rise to rapidly proliferating cells, which are called D- cells by Seri and colleagues (Seri et al., 2001; Seri et al., 2004). These cells have a completely different morphology than radial astrocytes. They have smooth plasma membranes, dark cytoplasm, many polyribosomes (more than astrocytes, less than granule cells), lighter mitochondria than astrocytes and an endoplasmatic reticulum that is larger than in astrocytes, but smaller than in granule cells. Three subtypes of D cells have been defi ned: D1, D2 and D3. D1 cells are small, mitotic cells with little cytoplasm and no extensions (Seri et al., 2004) or only a very thin process (Fukuda et al., 2003). D2 cells have a short thick process and D3 cells have characteristics of immature neurons, which are frequently branched 15

17 Chapter 1 dendrites that reach through the GCL into the molecular layer and a thin process projecting into the hilus. D2 and D3 cells are postmitotic (Seri et al., 2004). D cells stain positive with PSA-NCAM, doublecortin (DCX) and TUC-4, which are all proteins involved in neurite outgrowth, and for NeuroD, a transcription factor involved in neuronal differentiation. biii-tubulin, a marker for mature neurons, was only found in some D cells. Kempermann et al. (2004) use a different nomenclature and, based on different criteria, further distribute the proliferating precursor population into various subtypes. First, as described earlier, there is the type 1 GFAP positive cell that resembles the radial glia cell that can be found during development. According to his model, the remaining GFAP-negative cells of the proliferating population of precursors can be subdivided into type 2A (nestin-positive, DCX-negative), type 2B (nestin-positive and DCX-positive) and type 3 (nestin-negative, DCX-positive) cells. Type 3 cells are also positive for PSA-NCAM, though NCAM polysialylation also occurs frequently in type 2B cells (Kronenberg et al., 2003). Type 2 and type 3 cells look largely similar, but during the type2/type3 stage, cells undergo great morphological changes. Finally, maturation state of the different precursors can be defi ned based on their electrophysiological properties. Fukuda and colleagues (2003) showed that cells that express both nestin and PSA-NCAM (named type 2B by Kempermann et al. and type II by Fukuda et al.), can be distinguished from other cell types in the GCL on the basis of their high input resistance. Compared to adult mice (388 MΩ) (Van Praag et al., 2002), or adult rats ( MΩ) (Wang et al., 2000), the input resistance of these type II cells (532 MWΩ to 8.46 GΩ) can be considered extremely high. Cells with such a high input resistance were also defi ned by others and named class 1 cells (Ambrogini et al., 2004a). Moreover, class 1 cells do not respond to medial perforant path (mpp) stimulation, indicating that they do not receive input from this brain region. In addition, no spontaneous activity could be recorded in these cells. Cells that do respond to mpp stimulation are named class 2 cells, which can in turn be subdivided in four subclasses, based on their different responses to the stimulation Maturation of the newly formed cells After becoming postmitotic, newly formed cells start to express the neuronal marker NeuN and the calcium-binding protein calretinin. In this stage, cells are named type 4 cells (Kempermann et al., 2004). During further maturation, cells switch from expressing calretinin to calbindin. New neurons rapidly form axons towards the CA3 region. Fast Blue and FluoroRuby injections in the CA3 region at different time points after BrdU injection demonstrate that, in rats, newly generated neurons start to grow axons to the CA3 area between 4 and 10 days after their formation (Hastings and Gould, 1999; Stanfi eld and Trice, 1988). Analysis of newly formed neurons that had retrovirally been labeled with GFP during cell division revealed that 4-week old neurons still have smaller soma than 4-month old granule neurons. Moreover, they have a shorter total dendritic length, a lower number 16

18 General introduction of branch points and a lower spine density (Van Praag et al., 2002). It takes two to four weeks before newly formed neurons are functionally indistinguishable from other neurons (Van Praag et al., 2002). Four-week old neurons are immunoreactive for synaptophysin and have dendritic spines, indicating that they receive synaptic input (Markakis and Gage, 1999; Van Praag et al., 2002). In addition, newly generated neurons display spontaneous postsynaptic currents, with a fast onset (less than 1 ms) and a slower exponential decay (greater than 5 ms), which is typical of postsynaptic response to fast neurotransmitters, such as GABA. Furthermore, new neurons receive input from the perforant path and they show mostly similar electrophysiological properties compared to mature dentate cells, indicating that newly generated neurons are functionally similar to mature cells (Van Praag et al., 2002) Fluctuations in newly formed cell number The estimated cell cycle time of proliferating hippocampal progenitors is h for rats and h for mice. The length of the S-phase, during which BrdU can be incorporated into the DNA, is estimated to be 9-10 h for rats and 6-9 h for mice (Cameron and McKay, 2001; Hayes and Nowakowski, 2002). BrdU is thought to label cells within 2 h after injection. After a single injection with BrdU, the number of labeled cells roughly doubles between 2 h and 24 h (Cameron and McKay, 2001; Dayer et al., 2003; Palmer et al., 2000). In rats, the BrdU-positive cell number continues to increase at least until day 7 after the initial labeling (Palmer et al., 2000), or even until day 10 (Hastings and Gould, 1999). In mice, the BrdU-positive cell number rises until day 3 (Jessberger and Kempermann, 2003; Kempermann et al., 2004; Kronenberg et al., 2003; Steiner et al., 2004). After the peak in the number of newly generated cells has been reached, the amount of labeled cells strongly decreases. This decrease is thought to be caused by apoptosis of the newly formed cells or by label dilution resulting from ongoing proliferation (Prickaerts et al., 2004). The latter explanation is less likely, because between 2 h and 4 days after labeling with BrdU, the BrdU-positive cell population loses expression of the proliferation marker Ki-67, indicating that the cells exit the cell cycle (Dayer et al., 2003). In addition, the neurogenic regions in the brain has been shown to contain high numbers of apoptotic cells (Biebl et al., 2000). Between two and four weeks after cells have been generated, the number of BrdU-positive cells does not decrease any further and remains stable for at least months (Cameron and McKay, 2001; Dayer et al., 2003; Hastings and Gould, 1999; Kempermann et al., 2003; Kempermann et al., 2004; Palmer et al., 2000). The net number of newly generated cells in the DG per day is estimated to be 9000 in 10-week old rats (Cameron and McKay, 2001), which is approximately 1% of the total granule cell population. The newly formed cell number rapidly diminishes with age (see section 4.2). The rate of hippocampal neurogenesis shows a large variation between species. The macaque, for instance, only forms 200 new hippocampal cells per day, which represents only 0.004% of the granule cell layer (Kornack and Rakic, 1999). 17

19 Chapter 1 The fact that many of the newly formed cells die soon after they have been generated, implies that the timing of BrdU-injections with respect to the sacrifice of the animals crucially influences the interpretation of the data. Injections that are applied hours to a few days before termination will provide information on the proliferation of undifferentiated precursor cells. However, if animals are injected with BrdU one week or more in rats or three days or more in mice before they are sacrificed, BrdU staining will give an indication about the survival of newly generated cells and allow double labelings to study the phenotype of the BrdU-positive cells. Therefore, strategic injections with BrdU in combination with immunocytochemical stainings for proliferation markers, such as Ki-67, and immature neuron markers, such as DCX, will result in a relatively complete picture of the entire neurogenic process in one individual animal. 4. Regulation of hippocampal neurogenesis 4.1. The neurogenic microenvironment Adult hippocampal neurogenesis is thought to take place in an exclusive neurogenic niche. This idea is supported by the observation that transplantation of stem cells from the hippocampus or from the spinal cord into the hippocampus gives rise to new neurons, whereas transplantation to different parts of the CNS, such as the spinal cord, does not result in neuronal differentiation (Shihabuddin et al., 2000). A possibly important factor for the creation of a permissive microenvironment for neurogenesis is the vasculature. Proliferating clusters in the SGZ tend to be located around small capillaries. This close proximity to the vasculature appears to be exclusive for the DG, because proliferating precursors in other areas of the hippocampus do not show this close association with a vascular niche (Palmer et al., 2000; Palmer, 2002). There is also in vitro evidence for a role of blood vessels in promoting neurogenesis. Coculturing of NSC s with endothelial cells, but not with vascular smooth muscle cells, has been reported to promote NSC proliferation (Shen et al., 2004). In addition, the presence of endothelial cells delays differentiation of the NSC s and stimulates the formation of neurons. Since this effect can be accomplished without direct cell-cell contact between NSC s and endothelial cells, endothelial cells probably produce soluble factors that promote neurogenesis. A possible candidate is the vascular endothelial growth factor vegf, which also has been shown to promote cell proliferation (Cao et al., 2004; Schanzer et al., 2004; Wurmser et al., 2004). Next to the vasculature, hippocampal astrocytes are also likely to play a role in neuronal fate instruction of hippocampal precursors (Horner and Palmer, 2003; Song et al., 2002; Svendsen, 2002). When NSC s are differentiated in the presence of astrocytes derived from the adult hippocampus, the number of neurons increases tenfold (Song et al., 2002). Astrocytes from the adult hippocampus are about twice as effective as astrocytes that are obtained from the neonatal hippocampus in promoting neurogenesis. Astrocytes 18

20 General introduction from the adult spinal cord do not stimulate neurogenesis (Song et al., 2002), indicating that adult hippocampal astrocytes provides specifi c cues that promote neurogenesis Internal and external factors that regulate hippocampal neurogenesis There is a large variety of factors that have been shown to influence hippocampal neurogenesis. Age is perhaps the most important variable to affect neurogenesis. Hippocampal cell proliferation occurs at high rates in very young rats, but as the animals reach the youngadult stage, proliferation has reduced with approximately 75% and it further declines to hardly detectable levels in aged animals (Heine et al., 2004a; Kempermann et al., 1998; Kuhn et al., 1996). Stress and glucocorticoids have also repeatedly been reported to decrease hippocampal neurogenesis. Injections with corticosterone, for instance, have a negative effect on hippocampal cell proliferation, whereas adrenalectomy promotes the generation of new cells (Cameron and Gould, 1994). Also, acute psychosocial stress (Gould et al., 1997) or chronic unpredictable stress have been shown to reduce progenitor proliferation (Heine et al., 2004b) and survival (Westenbroek et al., 2004) of newly formed cells in male rats. A similar chronic stress protocol in female rats, however, caused an increased survival of newly generated neurons (Westenbroek et al., 2004). Besides corticosterone, other hormones can also affect hippocampal neurogenesis. Estrogen, for instance, has complicated effects on hippocampal cell proliferation. Administration of a high dose of estradiol in female meadow voles initially increases cell production, but reduces it again after 48 hours (Ormerod et al., 2003). In female rats, progenitor proliferation fluctuates throughout the estrus cycle, with highest levels during proestrus, but net neurogenesis is not different from males (Tanapat et al., 1999; Tanapat et al., 2005). Several neurotransmitter systems are also potent inhibitors or activators of hippocampal neurogenesis. Glutamatergic signaling via the NMDA receptor, for instance, has been reported to have a negative effect on progenitor proliferation (Cameron et al., 1995). Reduction of NMDA receptor activation by pharmacological blockage of the NMDA receptor or by damaging the major excitatory input into the hippocampus, the entorhinal cortex, stimulates cell proliferation and the formation of new neurons (Cameron et al., 1995; Nacher et al., 2001b). Another neurotransmitter that has been shown to have an impact on neurogenesis is serotonin or 5-HT. Pharmacological inhibition of serotonin synthesis or selective lesions of the serotonergic neurons strongly reduce hippocampal cell proliferation (Brezun and Daszuta, 1999). In addition, antagonists for the 5-HT1A and 5-HT2A receptor decrease the number of dividing cells (Banasr et al., 2001; Radley and Jacobs, 2002). 5-HT1A receptor agonist treatment, on the other hand or stimulation of serotonergic neurotransmission by chronic administration of antidepressants promote the formation of new cells (Banasr et al., 2001; Malberg et al., 2000). Hippocampal neurogenesis is also positively influenced by a variety of growth fac- 19

21 Chapter 1 tors and neurotrophic factors. Peripheral (Aberg et al., 2000) or central (Lichtenwalner et al., 2001) administration of insulin-like growth factor-1 (IGF-1), central infusion of basic fi broblast growth factor (bfgf) or epidermal growth factor (EGF) (Jin et al., 2003a) or overexpression of vascular endothelial growth factor (vegf) (Cao et al., 2004) activate progenitor proliferation. VEGF also has survival-promoting effects on newly formed cells (Schanzer et al., 2004). Moreover, BDNF has also been shown to have positive effects on the survival of newly formed neurons and the differentiation towards the neuronal phenotype (Sairanen et al., 2005; Scharfman et al., 2005). Furthermore, pathological conditions can affect adult hippocampal neurogenesis. Global and focal cerebral ischemia (Kee et al., 2001; Liu et al., 1998a; Tureyen et al., 2004; Zhu et al., 2004a), excitotoxic and mechanic lesions of the granule cell layer (Gould and Tanapat, 1997), or epileptic seizures (Cha et al., 2004; Hellsten et al., 2002; Parent et al., 1997; Sankar et al., 2000) all result in increased levels of hippocampal neurogenesis. Also in humans, there appears to be a link between brain pathology and neurogenesis, because the hippocampi of Alzheimer patients were reported to contain more cells that express markers for immature neurons than age-matched healthy controls (Jin et al., 2004). In animal models of ischemia, newly formed cells have been shown to migrate to the damaged area where some of them differentiate into neurons and may contribute to repair (Arvidsson et al., 2001; Jin et al., 2003b; Nakatomi et al., 2002) Effects of physical exercise on hippocampal neurogenesis Part of this thesis will be focused on the relation between running wheel activity and hippocampal neurogenesis. Physical activity has repeatedly been shown to cause a robust enhancement of hippocampal cell proliferation and the number of newly formed neurons. Mice that are housed with a running wheel show a large increase in the proliferation of hippocampal progenitors and the production of new granule neurons (Van Praag et al., 1999a; Van Praag et al., 1999b). Also in rats, beneficial effects of running on hippocampal neurogenesis have been demonstrated. Female spontaneous hypertensive rats (SHR), like mice, show voluntary running behavior when their cage is equipped with a wheel, which increases the production of new neurons with approximately 500% (Persson et al., 2004). Also non-voluntary running in a treadmill can promote hippocampal neurogenesis in rats of different ages (Kim et al., 2004; Trejo et al., 2001), but only if the intensity level of the exercise is mild to moderate (Kim et al., 2003). Running-induced stimulation of cell proliferation is restricted to the early stages of granule cell development, when cells express nestin. Radial glia proliferation remains unaffected, but proliferation of type 2A and type 2B cells (Fig. 1) is increased. Mitotic cells which have already lost nestin expression remain unaffected by exercise (Kronenberg et al., 2003). Exercise only affects hippocampal neurogenesis and does not influence neuron formation in the other neurogenic region of the brain, the olfactory bulb (Brown et al., 2003a). 20

22 General introduction 5. Neurogenesis and learning Despite the increasing number of studies that have been performed during the last decades to unravel the function of neurogenesis in the adult hippocampus, it is still unclear why new neurons are being generated in this brain structure. The fact that hippocampal neurogenesis is a phenomenon that can be observed in a large variety of species, supports the hypothesis that newly generated neurons have a distinct role in hippocampal functioning. Because of the well-known role of the hippocampal formation in learning and memory processes, it is tempting to speculate on a potential role for newly formed neurons in learning. There is increasing experimental evidence that there is a relation between learning and neurogenesis, though the data that are obtained are not equivocal. Here, a summary will be provided showing the variety of studies that have been performed to explore the role of hippocampal neurogenesis in learning Environmental impact on hippocampal neurogenesis The fi rst studies that indicated a role for newly formed hippocampal neurons in learning and memory were carried out in wild-living animals. In black-capped chickadees (Parus atricapillus) (Barnea and Nottebohm, 1994), for instance, neurogenesis appears to be related to seasonal variation in spatial memory processing. The number of new neurons in these birds shows a seasonal fluctuation, with most new neurons being produced during fall, when birds have to retrieve the spots where they have hidden food hours to days earlier. Interestingly, captive birds show a similar seasonal fluctuation of neurogenesis, though the total rate of new neuron production is only half of the free-living chickadees. In contrast to the fi ndings in the chickadees, no seasonal variations in hippocampal neurogenesis were observed in another food-storing species, the eastern grey squirrel (Sciurus carolinensis) (Lavenex et al., 2000), which suggests that fluctuations in neurogenesis across the year may be specific for birds. However, there is evidence that also in mammals, the complexity of the environment may be a determining factor for the rate of hippocampal neurogenesis. A comparison between four species of wild-living rodents shows that the species with the largest territory (yellow-necked and long-tailed wood mouse, Apodemus spp.) display the highest level of hippocampal neurogenesis (Amrein et al., 2004). Also in laboratory animals, enriched housing causes almost a doubling of the number of new neurons in young-adult mice and rats or even a fi vefold increase in aged mice, without influencing proliferation of undifferentiated precursor cells (Kempermann et al., 1997; Kempermann et al., 1998; Kempermann et al., 2002; Nilsson et al., 1999). Moreover, environmental enrichment results in a better performance in the Morris water maze (Kempermann et al., 2002; Nilsson et al., 1999). The positive effect of an enriched environment on neurogenesis is restricted to hippocampal neurogenesis. The number of newly generated neurons in the olfactory bulb does not increase by enriched housing (Brown et al., 2003a). 21

23 Chapter Effects of learning on neurogenesis There is also more direct evidence that there is a relation between learning and hippocampal neurogenesis. It has been shown that hippocampus-dependent learning can promote hippocampal neurogenesis in adult rats (Gould et al., 1999a). Training of rats in the trace eyeblink conditioning task (TEC) or the Morris water maze (MWM), one week after an injection with BrdU, has been shown to result in a doubling of the number of newly generated neurons. The hippocampus-independent versions of TEC and MWM, delay eyeblink conditioning and cued water maze learning, respectively, did not alter hippocampal neurogenesis. Follow-up studies showed that the increased number of new neurons that are formed during TEC remain for months after learning, a time point when the hippocampus is not required anymore for the retrieval of trace memories (Leuner et al., 2003). Using another protocol, in which rats were trained in the MWM directly after BrdU injection and sacrificed ten days after the last training session, Ambrogini et al. (2000) also showed positive effects of spatial learning on neurogenesis, although those effects were restricted to the infrapyramidal blade of the dorsal granule cell layer. The same authors reported that training of rats in the MWM 8-10 days after BrdU injection caused a significant decrease in hippocampal neurogenesis (Ambrogini et al., 2004b). In another study in rats (Snyder et al., 2005), in which BrdU was injected one week before the start of training, and in mice (Van Praag et al., 1999b), in which BrdU was injected before and during MWM training, no learning-induced alterations in the number of newly generated neurons were observed. Not only cell survival can be influenced by learning, it has also been reported that hippocampal progenitor proliferation may increase during spatial learning in rats (Lemaire et al., 2000), though this phenomenon was not observed by others (Gould et al., 1999a; Van Praag et al., 1999b). There is also evidence for complex fluctuations in cell production throughout the learning process (Dobrossy et al., 2003). Cell proliferation has been shown to be stable during the early phase of Morris water maze learning, but to increase during the asymptotic phase of the learning curve. At the same time, cells that have been generated during the early phase of learning die during the later stages of the learning process. Therefore, the net level of cell production remains unaltered during learning. Moreover, there are data that indicate a positive correlation between progenitor proliferation and spatial learning capacity in aged rats (Drapeau et al., 2003), but this is contradicted by others (Bizon et al., 2004; Merrill et al., 2003). Associative learning tasks, such as fear conditioning and two-way active avoidance learning, were shown to reduce hippocampal cell proliferation (Malberg and Duman, 2003; Pham et al., 2005), whereas another associative task, TEC, had no effect on cell proliferation (Gould et al., 1999a). A summary of the studies that tested the effects of learning on hippocampal neurogenesis is provided in Table 1. 22

24 General introduction 5.3. Effects of reduced neurogenesis on learning A different approach to investigate the relation between hippocampal neurogenesis and learning is to block hippocampal cell proliferation and to look at the behavioral consequences. Two weeks of treatment with the non-specifi c mitosis-inhibitor methylazoxymethanol acetate (MAM), which reduces neurogenesis with approximately 80%, signifi - cantly impaired TEC and trace fear conditioning in rats (Shors et al., 2001; Shors et al., 2002). Partial reduction of neurogenesis is not detrimental for all types of hippocampusdependent learning. MAM treatment or brain irradiation, which also dramatically reduces hippocampal neurogenesis (Monje et al., 2002), had no effect on MWM learning (Madsen et al., 2003; Shors et al., 2002). Moreover, contextual fear conditioning, another hippocampus-dependent learning task, remained unchanged after injections with MAM (Shors et al., 2002). Although the acquisition of many hippocampus-dependent tasks is not impaired by inhibition of neurogenesis, recent studies suggest that neurogenesis may be crucial for the consolidation of memories. Almost total ablation of hippocampal neurogenesis impaired long-term memory retention in the Morris water maze (Rola et al., 2004; Snyder et al., 2005) and it prevented the environmental enrichment-induced improvement in longterm recognition memory (Bruel-Jungerman et al., 2005) Computational models of neurogenesis: neurogenesis and memory clearance Computational models have been used to predict the role of hippocampal neurogenesis in learning and memory. In a simple model of a three-layered network without backward or lateral signaling, in which new cells are randomly inserted into the middle layer and old cells are randomly deleted, removal of old memories is strongly accelerated. Moreover, acquisition of new memories is accelerated in the model containing neurogenesis, especially with increasing network activity (Chambers et al., 2004; Deisseroth et al., 2004). There is also in vivo evidence for al role for neurogenesis in memory clearance (Feng et al., 2001). Presenilin-1 conditional knockout mice do not show an increase in hippocampal neurogenesis upon cage enrichment. When these transgenic mice were trained in a contextual fear conditioning paradigm and subsequently housed in enriched conditions for two weeks, retention was signifi cantly better than in control mice in which neurogenesis was increased during the enrichment procedure. Without enrichment, both mouse lines performed similarly in the retention test. These data suggest that the increase in neurogenesis in the wildtype mice after enrichment removed old memories from the hippocampus. 23

25 Chapter 1 Table 1: Overview of studies exploring the eff ects of learning on hippocampal neurogenesis. Abbreviations: ASA: Active shock avoidance, CFC: Contextual Fear Conditioning, DEC: Delayed Eyeblink Conditioning, MWM: Morris Water Maze, TEC: Trace eyeblink conditioning. Publication Species, strain Age/weight Learning task Protocol Effect on BrdUpositive cell number Gould et al. (1999a) Male rat, Sprague- Dawley g MWM/ TEC BrdU (200 mg/kg): day -7 Training: between day 1 and 8 Sacrifice 24 h after training No change Gould et al. (1999a) Male rat, Sprague- Dawley g MWM/ TEC BrdU (200 mg/kg): day -7 Training: between day 1 and 8 Sacrifice 7 d after training Increase Gould et al. (1999a) Male rat, Sprague- Dawley g Cued MWM/ DEC BrdU (200 mg/kg): day -7 Training: between day 1 and 8 Sacrifice 7 d after training No change Van Praag et al. (1999b) Van Praag et al. (1999b) Female mouse, C57Bl/6 Female mouse, C57Bl/6 3 months MWM BrdU (50 mg/kg): day 1 to 12 Training: day 1 to 23 Reversal training: day 24 to 30 Sacrifice: day 40 3 months MWM BrdU (50 mg/kg): day 1 to 12 Training: day 1 to 12 Sacrifice: day 13 No change No change Ambrogini et al. (2000) Male rat, Sprague- Dawley 2 months MWM BrdU (50 mg/kg): day -3 to -1 Training: day 1 to 5 Sacrifice: day 15 Increase (infrapyramidal bade) Lemaire et al. (2000) Male rat, Sprague- Dawley 4 months MWM BrdU (unknown conc.): day 3 to 5 (before first trial) Training: day 1 to 5 Sacrifice: day 6 Increase 24

26 General introduction Table 1, continued Publication Species, strain Dobrossy et al. (2003) Male rat, Sprague- Dawley Dobrossy et al. (2003) Male rat, Sprague- Dawley Dobrossy et al. (2003) Male rat, Sprague- Dawley Malberg and Duman (2003) Male rat, Sprague- Dawley Ambrogini et al. (2004b) Male rat, Sprague- Dawley Snyder et al. (2005) Male rat, Long Evans Pham et al. (2005) Male rat, Sprague- Dawley Age/weight Learning task Protocol Effect on BrdUpositive cell number 2 months, 200 g MWM BrdU (50 mg/kg): day 1 to 8 (before first trial) Training: day 1 to 8 Sacrifice: day 9 No change 2 months, 200 g MWM BrdU (50 mg/kg): day 5 to 8 (before first trial) Training: day 1 to 8 Sacrifice: day 9 Increase 2 months, 200 g MWM BrdU (50 mg/kg): day 1 to 4 (before first trial) Training: day 1 to 8 Sacrifice: day 9 Decrease g ASA BrdU (100 mg/kg): directly after training Training: 30 trials on 1 day Sacrifice: 2 h after BrdU Decrease 5 months MWM BrdU (50 mg/kg): day -10 to -8 Training: day 11 to 15 Sacrifice: day 18 Decrease ~2 months MWM BrdU (200 mg/kg): day -7 Training: day 1 to 6 Sacrifice: 1, 2 or 4 w after training No change 3 months CFC BrdU (200 mg/kg): immediately after CS exposure Training: pre-exposure to CS on day 1, exposure to CS + shock on day 2 Sacrifice: 2 h after BrdU Decrease 25

27 Chapter 1 6. Outline of this thesis The fi rst part of this thesis describes various factors that may influence hippocampal neurogenesis. In chapter 2, we investigated baseline fluctuations in hippocampal cell proliferation in mice. We assessed whether the cell cycle of hippocampal progenitors follows a circadian rhythm. Furthermore, this chapter describes the effects of acute sleep deprivation during the resting phase of the animals on the number of proliferating progenitors. Third, we studied if physical exercise, which mainly takes place during the active dark phase of the mice, induces a synchronization of the proliferating cell population. In chapter 3, the effects of exercise on hippocampal cell proliferation were investigated in more detail. In this chapter, we were interested in the temporal dynamics of the exercise-induced increase in neurogenesis. Are these effects acute or is long-term exercise required? And how long do the effects on hippocampal cell proliferation persist after removal of the running wheel from the cage? Chapter 4 describes an experiment in which we investigated the role of the medial septum in hippocampal neurogenesis. The medial septum forms one of the main input pathways to the hippocampus and it is an important brain structure for learning and memory. In addition, it is known that during physical activity, the septohippocampal connection is strengthened. Therefore, the medial septum can be considered as a possible mediator in the effects of learning or exercise on hippocampal neurogenesis. The studies described in the next three chapters investigate the potential role of hippocampal neurogenesis in learning and memory processes. Chapter 5 describes an experiment in which rats were exposed to either one or four days of active shock avoidance learning. These data provide insight into the effects of an associative learning task, which is hippocampus-independent, but which has been shown to activate cells in the dentate gyrus, on hippocampal cell proliferation and survival of newly formed cells. Chapter 6 explores the effects of a spatial learning paradigm, Morris water maze learning, on different aspects of hippocampal neurogenesis and plasticity. The study was performed with two frequently used strains, Wistar and Sprague-Dawley. Chapter 7 also describes experiments related to spatial learning and neurogenesis. Using the Y-maze as a learning task, we investigated whether an increase in the number of newly formed neurons would increase the rate of acquisition in a spatial learning task. Neurogenesis was enhanced by housing animals with a running wheel for 14 days. Secondly, we tested the hypothesis that enhanced neurogenesis following acquisition results in the clearance of information from the hippocampus. Third, we explored potential changes in hippocampal neurogenesis during memory retrieval, when the hippocampal memory trace is reactivated. 26

28 General introduction 27

29 28

30 2 Chapter Hippocampal cell proliferation across the day: Increase by running wheel activity, but no effect of sleep and wakefulness Karin van der Borght 1, Francesca Ferrari 1,2, Karin Klauke 1, Viktor Roman 1, Robbert Havekes 1, Andrea Sgoifo 2, Eddy A. van der Zee 1, Peter Meerlo 1 1) Department of Molecular Neurobiology, Graduate School of Behavioral and Cognitive Neurosciences, University of Groningen, The Netherlands; 2) Department of Evolutionary and Functional Biology, Stress Physiology Lab, University of Parma, Italy. Behavioural Brain Research (2005), in press 29

31 Chapter 2 Abstract The present study investigated whether proliferation of hippocampal progenitors is subject to circadian modulation. Mice were perfused using 3-h intervals throughout the light-dark cycle and brains were stained for Ki-67. Since Ki-67 is not expressed during the G0 phase of the cell cycle, we expected a decline in Ki-67 expression at the moment cells synchronously exit the cell cycle. However, despite the fact that various hippocampal factors fl uctuate across the day, the number of dividing cells remained constant. In a second experiment, we studied whether disturbance of normal sleep aff ected the stable rate in cell proliferation. Our data show that twelve hours of sleep deprivation during the light phase did not infl uence proliferating cell number. A third experiment investigated whether physical activity, a condition known to enhance hippocampal cell proliferation, caused an elevation of the steady baseline number of proliferating progenitors, or a peak directly following the active phase of the animals. Mice were housed with a running wheel for 9 days. On the last day, animals were sacrifi ced either directly before or directly after the active phase. Exercise signifi cantly promoted cell proliferation and this eff ect appeared to be strongest directly after the active period and to disappear during the resting phase. Our data suggest that hippocampal cell proliferation is not synchronized under basal conditions and is unchanged by sleep deprivation. However, running aff ected cell proliferation diff erentially at two times of day. These data demonstrate that the steady rate in cell proliferation is not indispensable, but can be changed by behavioral activity. 30

32 Hippocampal cell proliferation across the day Introduction The dentate gyrus of the adult hippocampus contains undifferentiated, rapidly proliferating progenitor cells. Approximately 70-80% of the newly formed cells differentiate into granule neurons, which ultimately fully integrate into the hippocampal network (Cameron et al., 1993; Kaplan and Hinds, 1977; Van Praag et al., 2002). The function of the neurons that are formed during adulthood is unclear, although data suggest that they play a role in the regulation of mood (Kempermann, 2002; Malberg et al., 2000; Santarelli et al., 2003) or in hippocampus-dependent (Gould et al., 1999a; Shors et al., 2001; Snyder et al., 2005) learning and memory. In order to investigate the potential role of newly formed hippocampal granule neurons in normal brain function, the identification of factors that influence neurogenesis may be of crucial importance. In the present study, we investigated basal levels of hippocampal cell proliferation across the day. For various tissue types, such as oral mucosa, the intestinal mucosa, the kidney or the bone marrow, a partly synchronized cell cycle has been observed (Bjarnason et al., 1999; Burns et al., 1972; Scheving et al., 1978). This means that around the same time of day, a large portion of the dividing cells enters the same phase of the cell cycle. Studies in invertebrates suggest that the generation of new neurons may also fluctuate across the day. Neurogenesis in the olfactory pathway of a certain crustacean (the American lobster, Homarus americanus) shows diurnal variations with a peak in cell production around dusk (Goergen et al., 2002). This is the time of day when these crustaceans display the highest levels of activity. Also for rodents it has been suggested that there is an activity-mediated circadian rhythm in hippocampal cell proliferation (Holmes et al., 2004). We hypothesized that hippocampal cell proliferation varies across the day-night cycle as a consequence of daily rhythms in physiological parameters and neuroendocrine factors (e.g., neurotrophic factors or glucocorticoid hormones) or by spontaneous day-night differences in behavior (e.g., the daily rhythm in sleep and wakefulness and variations in activity level across the day). In addition, external factors, such as the light-dark cycle, and the consequent daily rhythm in light exposure also may affect hippocampal function and the number of proliferating progenitors. In order to test this hypothesis, mice were sacrificed throughout the day using 3-h intervals. Brains were stained for the proliferation marker Ki-67. Since Ki-67 is expressed during all phases of the cell cycle, except G0 (Scholzen and Gerdes, 2000), a decline in Ki- 67 expression at a certain time of day would be expected at the time that a large number of cells concurrently fi nish the cell cycle (Bjarnason et al., 1999). We avoided the use of BrdU, since this substance has to be administrated via injections, which would arouse the animals and disturb their sleep and activity pattern. Here we show that, despite the fact that many hippocampal processes fluctuate during the day, there is no evidence for synchronized hippocampal cell proliferation under a normal sleep-wake cycle. Next, we investigated whether the stable rate in proliferating cell number is affected by disturbance of normal sleep or by increased physical activity. To determine the role of sleep in the regulation of hippocampal cell proliferation, we sleep deprived mice by 31

33 Chapter 2 gentle handling during their normal resting phase and determined the number of proliferating cells in the dentate gyrus. The impact of increased physical activity on the constant generation of new hippocampal cells across the day was studied by housing animals with a running wheel for 9 days. Wheel running has repeatedly been show to enhance the number of proliferating cells in the hippocampus (Holmes et al., 2004; Van Praag et al., 1999b). Here, we explored whether running wheel activity causes an increase in cell proliferation that is constant across the day or whether increased activity results in an acute peak in proliferating cell number, which declines during the resting phase. In order to answer this question, mice were either sacrificed directly following the active phase or at the end of the resting phase, two time points at which the rate of hippocampal cell proliferation is identical under basal conditions. The results suggest that the stable rate of cell proliferation is not affected by sleep deprivation, but may be enhanced during the active phase, by high levels of physical activity. Materials and Methods All experiments were performed with male C57Bl/6 mice (Harlan, Horst, The Netherlands). Animals were individually housed under a 12h light/12h dark cycle and were provided with food and drinking water ad libitum. The procedures concerning animal care and treatment were in accordance with the regulations of the ethical committee for the use of experimental animals of the University of Groningen. Experiment 1: Spontaneous daily fl uctuations in hippocampal cell proliferation To investigate spontaneous fluctuations in hippocampal cell proliferation throughout the light-dark cycle, week-old mice were sacrificed using 3-h intervals (n=5-8 per time point), starting 2.5 h after the beginning of the light period (Zeitgeber Time (ZT) 2.5, with light onset designated as ZT 0). Animals were rapidly anesthetized with CO 2. Subsequently, mice were transcardially perfused with heparinized saline followed by 4% paraformaldehyde in 0.1 M phosphate buffer. Experiment 2: The eff ect of sleep deprivation on hippocampal cell proliferation The effects of sleep on cell proliferation in the dentate gyrus were studied by depriving 6-8-week-old mice (n=7) of sleep for h starting at the onset of the light phase, the period during which the mice normally sleep most of the time. Sleep deprivation was accomplished by the gentle handling method, which involved tapping on the cage, gently shaking the cage or, when this was not suffi cient to keep the animals awake, disturbing the sleeping nests. Control animals (n=8) were left undisturbed. During the sleep deprivation period, the experimenter recorded number and nature of the interventions that were necessary to keep the animals awake. After 8-10 h of sleep deprivation, mice were injected i.p. with 300 mg/kg BrdU (Sigma, St. Louis, MO, USA) dissolved in saline (20 mg/ml, ph 7.0). Two hours after the BrdU injection, animals were perfused. Prior to perfusion, a 32

34 Hippocampal cell proliferation across the day blood sample (50 μl) was taken from the heart. Blood samples were collected in prechilled Eppendorf tubes containing EDTA as anti-coagulant. After centrifugation at 2600 rpm for 15 min, the plasma was collected, and stored at -80 C for later analysis of corticosterone by radioimmunoassay (ICN Biomedicals, Costa Mesa, CA, USA). Experiment 3: The eff ect of exercise on hippocampal cell proliferation To assess whether day-night variations in activity level can induce synchronization of hippocampal cell proliferation, 6-8-week-old mice (n=15) were housed with a running wheel for 9 days to increase physical activity. Control animals (n=16) were housed in a standard cage during this period. On the last day of the exercise period, 8 runners and 8 control animals were sacrificed just before the active phase (ZT 10-12). The remainder of the animals was perfused shortly after the circadian activity phase (ZT 0-2). Brain processing and immunocytochemistry Following the perfusions, brains were removed from the skull and kept in 0.01 M PBS overnight. Subsequently, the brain material was cryoprotected in 30% sucrose in 0.1 M phosphate buffer for 48 h and 30 μm sections were cut with a cryostat. Sections were kept in 0.01 M PBS containing 0.1% sodium-azide until further processing. For both BrdU and Ki-67 immunocytochemistry, every sixth section of the dorsal hippocampus was used. The material for the BrdU immunostaining underwent steps for DNA denaturation (van der Borght et al., 2005a). Sections were placed in a 50% formamide/2xssc solution at 65 C for 20 min, followed by a rinsing step with 2XSSC. Next, sections were incubated with 0.2 M HCl at 37 C for 30 min and neutralized in a 0.1 M borate buffer solution. Sections for the BrdU as well as for the Ki-67 staining were incubated with 0.3% H 2 O 2 for 30 min. This was followed by 3% normal serum (Jackson Immunolabs, West Grove, PA, USA) and 0.1% Triton-X100 in 0.01 M PBS. The primary antibody (rat-anti-brdu, 1:800, Oxford Biotechnology, Oxfordshire, UK or rabbit-anti- Ki-67, 1:400, Novocastra, Newcastle upon Tyne, UK) was applied for 48 h at 4 C. After a second blocking step with 3% normal serum and 0.1% Triton-X100, the biotinylated secondary antibody (donkey-anti-rat or goat-anti-rabbit, both 1:400, Jackson Immunolabs, West Grove, PA, USA) was added for 2 h. This was followed by incubation with Avidin- Biotin-Complex (1:400, ABC Elite kit, Vector Laboratories, Burlingame, CA, USA) for 2 h. The staining was visualized by adding 0.2 mg/ml diaminobenzidine (DAB) and 0.003% H 2 O 2. Quantifi cation BrdU or Ki-67 immunopositive cells were counted in every sixth section of the dorsal hippocampus using a 40x objective. Only cells that were located in the subgranular zone, or one cell diameter deviating from this region, were included. Cells throughout the entire thickness of the section were counted. The number of counted cells was multiplied by 6 to get an estimation of the total cell number in the dorsal dentate gyrus (Holmes et al., 2004; van der Borght et al., 2005b). 33

35 Chapter 2 #Ki6-7 postiv i e cel ls A B GC L Zeitgeber time H Figure 1: Hippocampal cell proliferation across the day. A) Groups of mice were sacrifi ced at 8 time points during the day and brains were stained for Ki-67. The white bar represents the light phase, the dark bar the dark phase. There were no signifi cant diff erences in proliferating cell number in the subgranular zone between the diff erent time points. B) Photomicrograph showing Ki-67 positive cells in the subgranular zone of the dentate gyrus. Scale bar = 50 μm. C) Magnifi cation of the selected area in the inner blade. Scale bar = 10 μm. GCL: Granule cell layer, H: hilus. C Statistics Circadian fluctuations in Ki-67 positive cell number were analyzed using a one-way ANO- VA. Effects of exercise on time-of-day variations in cell proliferation were tested with a two-way ANOVA. The sleep deprivation data were analyzed with an independent samples t-test. All data are presented as mean ± S.E.M. 34

36 # Interventio ns/h Hippocampal cell proliferation across the day A sounds cage shaking nest disturbance combined stimulation Continuous stimulation i-67 c ells B Duration of sleep deprivation (h) 0 control SD C D GCL H H GCL rd - o t s # B U p si ive cell E control SD Figure 2: Eff ect of h of sleep deprivation by gentle handling on hippocampal cell proliferation. A) Animals were prevented from falling asleep by sounds (tapping on the cage), carefully shaking the cage or disturbance of the sleeping nests. The number of interventions that was necessary to keep the animals awake gradually increased during the day. During the last 4 h, more than 60 interventions per hour were required to prevent the animals form falling asleep. No accurate numbers are available, because observations were interrupted by BrdU injections during the 9th and 10th hour and perfusions during the 11th and 12th hour of the sleep deprivation period. B) The sleep deprivation procedure had no eff ect on the number of Ki-67 positive cells in the subgranular zone. C) Representative photomicrograph showing BrdU labeled cells in the dentate gyrus. Scale bar = 100 μm. D) Magnifi cation of the selected area in C. Scale bar=30 μm. GCL: Granule cell layer, H: hilus. E) BrdU, which had been injected 2 h prior to sacrifi c- ing the mice, was incorporated in the same number of cells in sleep deprived and control mice. Results No daily rhythm in hippocampal cell proliferation In order to determine whether hippocampal cell proliferation shows fluctuations across the day, groups of mice were sacrificed throughout the light-dark cycle, using 3-h intervals. The number of proliferating cells was determined by counting Ki-67 immunoreactive nuclei in 35

37 # Ki-67 pos itive cells Chapter sedentary exercise * ZT 0-2 ZT Time of sacrifice Figure 3: Eff ect of running wheel activity on cell proliferation, measured at two diff erent times of the day. Cell proliferation in the subgranular zone was signifi cantly increased in animals that had access to a running wheel. An independent samples-t-test showed that this eff ect was only signifi cant (*, P<0.05) in animals that were sacrifi ced at the beginning of the resting phase at ZT 0-2. the dentate gyrus. At all time points investigated, the subgranular zone contained between 1500 and 1800 Ki-67 positive cells. There were no significant differences between the different time points that were investigated (Fig. 1A). No eff ects of sleep deprivation on hippocampal cell proliferation The influence of sleep on hippocampal progenitor proliferation was investigated by sleep depriving animals for h during the light phase. Although the mice were disturbed more frequently as the sleep deprivation period progressed, plasma levels of the stress hormone corticosterone at the end of the experiment were not elevated above control levels, suggesting that the animals were not seriously stressed (control: 11.0 ± 3.5 μg/dl, sleep deprived: 16.4 ± 4.4 μg/dl, P>0.05). Yet, the number of interventions that was necessary to keep the animals awake gradually increased during the sleep deprivation period, thereby indicating an increasing drive for sleep (Fig. 2A). Despite the accumulated sleep debt, neither the number of Ki-67 positive cells (Fig. 2B) nor the number of cells that had incorporated BrdU (Fig. 2C-E) was changed by the sleep deprivation procedure. Exercise stimulates cell proliferation, with the largest eff ect directly after the active period Animals were housed with a running wheel for 9 days and on the last day they were sacrificed at either ZT 0-2 or at ZT Animals from both groups ran on average the same distance over the 9-day exercise period (73.4 ± 7.2 km and 69.6 ± 6.3 km, respectively). Brains were stained for Ki-67 to study the effect of an increased activity level on hippocampal cell proliferation and to determine whether this effect is different directly after the active period compared to directly before the start of the dark phase. The data show that exercise caused a significant increase in cell proliferation in the subgranular zone (Fig. 3, twoway ANOVA, F(1,31)=5.63, P<0.05). There was no statistically significant effect of time of sacrifice or an interaction between exercise and the time of sacrifi ce. However, when tested separately with an independent-samples t-test, the effect of exercise on cell proliferation was only significant at the ZT 0-2 time point (P<0.05) and not at ZT

38 Hippocampal cell proliferation across the day Discussion The present study shows that, under standard housing conditions, hippocampal cell proliferation in the mouse does not show a circadian rhythm. In addition, the stable number of proliferating cells in the dentate gyrus is not affected by sleep deprivation during the resting phase of the animal. However, providing the mice with a running wheel significantly promoted hippocampal cell proliferation. This effect was strongest directly following the period of highest activity and seemed to disappear in the course of the resting phase. Circadian regulation of cell proliferation has been reported for many tissue types (Bjarnason et al., 1999; Burns et al., 1972; Scheving et al., 1978; Smaaland, 1996). A clear example of synchronous circadian cell cycling can be found in the liver. Upon ablation of a large part of the liver, resting hepatocytes enter the cell cycle in order to repair the damaged tissue. However, the time point at which the hepatocytes undergo transition from G2 to mitosis is strictly limited to a specifi c time of day, independent of the time of day at which the lesion was made (Matsuo et al., 2003). In contrast to these cell types, our data indicate that under standard laboratory conditions proliferation in the hippocampus is not subjected to circadian modulation. This is in line with previous studies, in which BrdU was injected at ZT 6, ZT 12 and ZT 18 (Holmes et al., 2004) or at ZT 1, ZT 7, ZT 13 and ZT 19 (Ambrogini et al., 2002) and no differences in BrdU-positive cell number were observed. Importantly, the lack of a circadian rhythm in cell proliferation after BrdU injections in these studies may have been caused by the fact that the animals were aroused and disturbed by the injections, which could have masked an endogenous rhythm. Therefore, we counted the number of Ki-67 positive cells in the dentate gyrus, which is also a reliable marker for dividing cells (Kee et al., 2002; Scholzen and Gerdes, 2000) and which does not require disturbance of the animals. Since Ki-67 is present throughout all phases of the cell cycle, fluctuations in Ki-67 expression indicate the simultaneous exit of cohorts of cells from the cell cycle. For other tissues, Ki-67 has been shown to be a suitable marker for observing circadian rhythms in cell proliferation (Bjarnason et al., 1999). However, in the hippocampus no differences were observed between the 8 time points investigated, suggesting that under basal conditions hippocampal cell proliferation is constant across the 24-h light-dark cycle. We further hypothesized that circadian fluctuations at the behavioral level (i.e. the sleep-wake rhythm and an uneven distribution of activity across the day) might influence the number of dividing cells at a certain time of day. To test whether disturbance of the normal sleep-wake pattern of the animals affects the steady daily rate in hippocampal cell proliferation, we sleep-deprived mice by gentle handling during their normal resting phase. The number of interventions that were required to keep the mice awake gradually increased in the course of the sleep deprivation period, illustrating an increasing sleep drive. With a single injection of BrdU, we labeled cells during the last two hours of the sleep deprivation period, but no effect of sleep loss on cell proliferation was observed. The brains were also analyzed for the number of Ki-67 expressing cells, because this method may provide 37

39 Chapter 2 information on the cumulative effects of h of sleep deprivation. However, despite the rigorous stimulation and the increasing sleep debt, also this staining did not show alterations in the number of dividing cells. Our data do not confi rm the fi ndings reported by Guzmán-Marín and colleagues (2003) showing that sleep deprivation in rats signifi cantly reduced cell proliferation. However, their study shows essential differences compared to the protocol we used. Besides the fact that we performed our experiment with mice instead of rats, the duration of the sleep deprivation period (12 h versus 96 h) and the method that was used to sleep deprive the animals (gentle handling versus the treadmill method) also differed between the two studies. Importantly, in the present experiment, we only counted proliferating cells in the subgranular zone, the brain region that is known to be the place were the neuronal progenitors are located (Alvarez-Buylla and Lim, 2004; Seki, 2002b). The cell counts reported by Guzmán-Marín and colleagues did not distinguish between the the subgranular layer and the hilus, which is known to be a more gliogenic region (Liu et al., 1998b; Picard-Riera et al., 2004; Rietze et al., 2000). Perhaps, the changes in cell proliferation that were reported in rats merely reflect changes in the hilar region and not in the subgranular zone. Indeed, a study in our lab showed that acute sleep deprivation in rats suppressed cell proliferation primarily in the hilus (Roman et al., 2005). From our study we can conclude that the amount of time that mice normally spend sleeping during the light phase is not of crucial importance for the regulation of cytogenesis in the granule cell layer. In a third experiment, we investigated the impact of increased physical activity on hippocampal cell division. We confi rmed previous reports in the literature, which did not take time-of-day into consideration, by showing that 9 days of running wheel activity increased hippocampal cell proliferation. We further investigated whether under conditions of increased voluntary exercise, cell proliferation was constantly enhanced across the day, or whether increased activity during the dark phase would alter the stable rate in cell proliferation and cause a peak immediately following a night of running. We therefore sacrificed groups of mice at two times of day, directly following the active phase and at the end of the resting phase. Although there is no statistically significant interaction between exercise and the time of day at which the animals were sacrifi ced, our data suggest that the exercise-induced increase in cell proliferation is most profound directly after the active phase and disappears in the course of the resting phase. This fi nding implies that exercise may acutely stimulate resting progenitors to enter the cell cycle, resulting in an accumulation of proliferating cells during the active period. The increased cell proliferation appears to be somewhat lower at the end of the resting phase. Possibly, as the estimated cell cycle time of hippocampal progenitors in mice is h (Hayes and Nowakowski, 2002), many of the cells that entered the cell cycle during the active phase completed the cell cycle before the end of the resting phase. Our data suggest that running wheel activity may not simply cause a continuously elevated number of dividing cells. It appears more likely that running induces certain factors that acutely promote cell proliferation. IGF-1 and vegf may be potential candidates responsible for this effect since they have been shown to increase cell proliferation under baseline condi- 38

40 Hippocampal cell proliferation across the day tions (Anderson et al., 2002; Schanzer et al., 2004) and to be essential for the exercise-induced increase in neurogenesis (Fabel et al., 2003; Trejo et al., 2001). However, additional research is required to determine the temporal dynamics of the exercise-induced upregulation of these two growth factors and of the effects of IGF-1 and vegf on promoting cell proliferation. A fi nal remark is that the number of Ki-67 positive cells considerably differed between the 3 independent experiments. This could be caused by the fact that experiments were performed with animals from different batches or by the fact that the mice of experiment 1 were a few weeks older than the other groups (Heine et al., 2004a; Kempermann et al., 2002). However, since we compared the experimental groups with the appropriate control groups, the difference in proliferating cell number between the experiments does not interfere with the interpretation of the data. In summary, our data show that under baseline laboratory conditions, hippocampal cell proliferation takes place at a constant rate across the day, which is not affected by 12 h of sleep deprivation during the resting phase. Despite the presence of hippocampal daily rhythms in various physiological parameters and in sleep-wake behavior, hippocampal cell proliferation remains stable. However, housing animals with a running wheel, resulting in high levels of physical activity, appears to induce a partly synchronous cell cycle. These data demonstrate that the steady rate in cell proliferation may not be indispensable for optimal hippocampal functioning, but can be changed by behavioral activity. Acknowledgements We thank Anouk Funke, Lisa van den Hengel, Ingrid Nijholt and Jan Bruggink for their assistance with the sleep deprivation, the perfusions and the corticosterone assay. This work was supported by the Netherlands Organization for Scientific Research (NWO-Vernieuwingsimpuls; E.A.V.d.Z. (grant ) and P.M. (grant )). 39

41 40

42 3 Chapter Differential regulation of hippocampal cell proliferation and cell survival by physical exercise Karin Van der Borght 1, Karin Klauke 1, Bart J.L. Eggen 2, Eddy A. Van der Zee 1, Peter Meerlo 1 1) Department of Molecular Neurobiology, Graduate school of Behavioural and Cognitive Neurosciences, and 2) Department of Developmental Genetics, Groningen Biomolecular Sciences and Biotechnology Institute; University of Groningen, P. O. Box 14, 9750 AA, Haren, The Netherlands 41

43 Chapter 3 Abstract Physical exercise has repeatedly been shown to enhance hippocampal neurogenesis. In the present study, we examined the temporal dynamics of this exercise-induced eff ect. Mice were housed with a running wheel for 1, 3 or 10 days and the number of proliferating progenitors and immature neurons were determined with Ki-67 and doublecortin (DCX) immunocytochemistry, respectively. We show a near-signifi cant increase in progenitor proliferation after 3 days of wheel running. After 10 days of enhanced activity, both Ki-67 and DCX-positive cell number were signifi cantly higher than in sedentary controls. Importantly, the number of proliferating cells was more closely associated with the distance run on the last day than with the total number of days the wheel had been available. In a second experiment, we investigated whether the positive eff ects of exercise on neurogenesis persisted after activity levels went back to baseline. After 10 days of access to a running wheel, we removed the wheel from the home cage and neurogenic rate was determined after 1 and 6 days of wheel deprivation. Our data show that the number of proliferating cells returned to baseline levels within 1 day of wheel withdrawal. The number of immature neurons, on the other hand, remained elevated for at least 6 days after the wheel had been removed. These data suggest that hippocampal cell proliferation and cell survival are diff erentially regulated by exercise. Cell proliferation responds acutely to fl uctuations in activity level, whereas the eff ects of wheel running on cell survival are more persistent. 42

44 Exercise and neurogenesis Introduction Physical exercise has various positive effects on plasticity and health of the central nervous system. It facilitates learning and memory (Fordyce and Wehner, 1993; Laurin et al., 2001; Van der Borght et al., 2005b; Van Praag et al., 1999a) and may prevent the cognitive decline that is observed during aging (Lytle et al., 2004). In addition, treadmill running in rats reduces the size of an infarct after middle cerebral artery occlusion (Ding et al., 2004) and accelerates recovery from brain injury (Stummer et al., 1994; Yang et al., 2003). Recent studies even indicate that increased physical activity delays the formation of b-amyloid plaques in a transgenic mouse model for Alzheimer s disease (Adlard et al., 2005; Lazarov et al., 2005). In the present study, we focused on hippocampal neurogenesis, another form of hippocampal plasticity that is positively affected by exercise. Enhanced physical activity increases proliferation of hippocampal progenitors and promotes survival of newly formed granule neurons in the dentate gyrus (DG) (Brown et al., 2003a; Jessberger and Kempermann, 2003; Kronenberg et al., 2003; Van der Borght et al., 2005a; Van Praag et al., 1999b). The mechanism underlying the exercise-induced elevation of hippocampal neurogenesis is not fully understood, although it likely involves a complex cascade of multiple mediators that includes growth-factors such as vegf (Cao et al., 2004; Fabel et al., 2003) and IGF-1 (Carro et al., 2000; Trejo et al., 2001). In order to gain more insight into the possible mechanisms underlying the beneficial effects of physical activity on the formation of new neurons, we investigated the temporal dynamics of the exercise-mediated increase in hippocampal neurogenesis. First, we studied whether exercise acutely stimulates neurogenesis or whether prolonged periods of enhanced physical activity are required to promote the generation of new cells. Mice were housed with a running wheel for 1, 3 or 10 days and the number of proliferating cells and immature neurons in the dentate gyrus were determined by Ki-67 and doublecortin (DCX) immunocytochemistry, respectively. Next, we studied how long the positive effects of wheel running on neurogenesis persist after activity levels have returned to baseline. Mice were provided with a running wheel for 10 days, after which the wheels were removed from the cage. Subsequently, the rate of hippocampal cell proliferation and the number of newly formed neurons were examined on day 0, 1 or 6 after wheel removal. Our data indicate that running wheel activity has an acute stimulatory effect on cell proliferation. After removal of the wheel, progenitor proliferation immediately returned to baseline, whereas the number of immature neurons remained elevated for at least 6 days, suggesting that proliferation rate and cell survival are differentially regulated by exercise. 43

45 Chapter 3 Materials and Methods Animals and housing All experiments were performed with adult male 8-week-old C57Bl/6 mice (Harlan, Horst, The Netherlands). Animals were individually housed in temperature-controlled climate rooms under a 12/12h light/dark cycle (lights on at 8:00 a.m.). The mice had free access to water and food throughout the experiment. All procedures concerning animal care and treatment were approved by the ethical committee for the use of experimental animals of the University of Groningen. Experimental procedure In the fi rst experiment, mice were housed with a running wheel (diameter 13 cm) for 1, 3 or 10 days (n=8 per group). Sedentary control mice (n=6-8 per group) were kept in standard cages without a wheel. Running wheel activity was recorded by a computer system (ERS, Haren, The Netherlands). Animals were sacrificed during the fi rst 4h of the light phase, following the last night of running. In the second experiment, mice were housed with a running wheel for 10 days (n=18) or kept in a standard cage (n=8). From the 18 animals that had access to a running wheel, 6 were sacrifi ced directly following the 10-day exercise period. The other 12 were returned to a standard cage without a running wheel and sacrificed 1 or 6 days later (n=6 each). Brain processing and immunocytochemistry for Ki-67 and DCX After anesthesia with 0.4 ml of sodium-pentobarbital, mice were perfused with heparinized saline, followed by 4% paraformaldehyde in 0.1 M phosphate buffer. Brains were kept in 0.01 M PBS overnight and subsequently cryoprotected by 30% sucrose for 48 h. With a cryostat, 6 series of 30 μm sections spanning the entire hippocampus were cut and collected in PBS containing 0.1% azide. Brain sections were stained for Ki-67, a nuclear protein that is expressed during all phases of the cell cycle, except G0 (Scholzen and Gerdes, 2000), and for doublecortin (DCX), a marker that can be found in immature neurons (Couillard-Despres et al., 2005; Rao and Shetty, 2004). In brief, free-floating brain sections were treated with 0.3% H 2 O 2 for 30 min, which was followed by incubation with 3% normal serum and 0.1% TritonX-100. Next, sections were exposed for 72 h to polyclonal rabbit-anti-ki-67 (1:400, Novocastra, Newcastle upon Tyne, UK) or to polyclonal goat-anti-dcx (1:1000, Santa Cruz Biotechnology, Santa Cruz, CA, USA). After thorough rinsing, brain slices were treated with 3% normal serum and 0.1% TritonX-100, followed by 2-h incubation with biotinylated goat-anti-rabbit for Ki-67 or biotinylated rabbit-anti-goat for the DCX staining (1:400 for both antibodies, both from Jackson Immunolabs, West Grove, PA, USA). Subsequently, the avidin-biotin complex (ABC Elite kit, Vector Laboratories, Burlingame, CA, USA) was added for 2 h, after which the staining was visualized with 0.2 mg/ml diaminobenzidine and 0.003% H 2 O 2. 44

46 Exercise and neurogenesis day 3 days 10 days A # K i-67 p ositiv e c e lls sedentary exercise # *** B Dst i anc e (km) Day 0 1 day 3 days 10 days D day 3 days 10 days C GCL Hilus # Ki-6 7 posti i ve cells E 1000 R 2 =0.55 P< Distance (km) GCL Hilus Figure 1: Time-dependent increase in cell proliferation by exercise. A) Mice gradually increased the distance run per day. The graph shows the average distance run per day ± S.E.M. B) Three days of enhanced activity caused a near-signifi cant increase in cell proliferation (#, P 0.1). After 10 days of wheel access, the number of Ki-67 positive cells was signifi cantly elevated (***, P 0.001). The graph represents the average number of Ki-67 cells per dentate gyrus ± S.E.M. C) The number of proliferating cells in the subgranular layer signifi cantly correlated with the distance run during the last day. D Representative photomicrograph of Ki-67 positive cells in the dentate gyrus of a sedentary mouse. Ten days of wheel running caused an increase in hippocampal cell proliferation (E). GCL: granule cell layer. Scale bar = 100 μm. Analysis of the Ki-67 and the DCX staining Ki-67 positive nuclei were counted in the subgranular zone of every sixth section of the dorsal hippocampus with a 40x objective. Immunopositive cells that were located one cell diameter deviating from this region were also included. Cell counts were performed throughout the entire thickness of the section. The total number of counted cells was multiplied by 6, in order to get an estimation of the total number of proliferating cells in the DG of the dorsal hippocampus (Holmes et al., 2004; van der Borght et al., 2005a). DCX-immunoreactivity was measured by quantifying the density of DCX-positive dendrites in the granule cell layer in 4 sections per animal. The majority of newly 45

47 Chapter 3 D CX- posit ive d endrit es sedentary exercise * 1 day 3 days 10 days B A C Figure 2: Time-dependent increase in the number of immature neurons by exercise. A) Only after 10 days of wheel running, the density of DCX-positive dendrites in the granule cell layer was signifi cantly increased (*, P 0.05). The graph shows the average area of the granule cell layer that was covered with DCX-positive dendrites ± S.E.M. B and C show examples of DCX immunoreactivity in a sedentary mouse and after 10 days of exercise, respectively. GCL: granule cell layer. Scale bar = 50 μm. Hilus GCL Hilus GCL formed neurons posses only 1 primary dendrite that does not show any branching within the granule cell layer. The area of the granule cell layer that is covered with DCX-positive dendrites therefore provides a reliable indication of the number of DCX-positive immature neurons (Van der Borght et al., 2005b). Measurements of the DCX-positive dendrite density were performed with a computerized system (Leica Qwin, Rijswijk, The Netherlands). Two equally sized areas in the inner blade and two areas in the outer blade, covering the entire width of the granule cell layer, were delineated in each section. The total surface within the demarcated areas that was covered with DCX-positive dendrites was determined. An average value per section was calculated for each animal. Statistics Effects of exercise on Ki-67 or DCX expression after 1, 3 or 10 days of running wheel activity were analyzed with a two-way ANOVA with treatment (sedentary or running wheel) and duration of the treatment (1, 3 or 10 days) as between-subjects factors. If this revealed significant effects, pairwise comparisons were made with a post hoc LSD test. In order to see if differences in cell proliferation and immature neuron number were primarily due to individual differences in running intensity or to the duration of wheel access, a linear regression analysis was performed with the distance that was covered on the last day and the total duration of the treatment as independent variables. Analysis of effects of running and wheel withdrawal on cell proliferation and the number of immature neurons was done with a one-way ANOVA. If this revealed a significant difference, a post hoc LSD test was performed for pairwise comparison. 46

48 Exercise and neurogenesis Results Eff ects of 1, 3 and 10 days of exercise on hippocampal neurogenesis To establish the temporal dynamics of the neurogenesis-promoting effects of running wheel activity, mice were housed with a running wheel for 1, 3 or 10 days. The animals ran on average 5 to 6 km per day during the fi rst three days, but in the 10-day exercise group this distance gradually increased up to an average of 13 km during the last day (Fig. 1A). Ki-67 immunocytochemistry was used to determine the rate of hippocampal cell proliferation (Fig. 1B, D and E). A two-way ANOVA revealed a significant effect of exercise (F(1,46)=12.23, P 0.001) and a significant interaction between the duration of wheel access and exercise (F(2,46)=7.99, P 0.001). A post hoc LSD test showed that 1 day of exercise was not suffi cient to enhance the rate of hippocampal cell proliferation. After 3 days of running wheel activity, a near-significant increase in Ki-67 positive cell number was observed (P 0.1). Ten days of running wheel access signifi cantly enhanced the number of proliferating cells (P 0.001). Linear regression analysis showed the duration of running wheel access and the distance run during the last day of the experiment together accounted for 53% of the variation in Ki-67 positive cell number (Fig. 1C, R2=0.53, F(2,23)=11.72, P 0.001). However, the variation in Ki-67 positive cell number in this model was largely related to variation in distance on the last day (P 0.05), but not to the duration of wheel access (n.s.). We also quantified the density of DCX-positive dendrites in the granule cell layer as a measure for the number of newly formed neurons (Fig. 2A-C). A two-way ANOVA showed a significant effect of the duration of running wheel exposure on the number of immature neurons (F(2,46)=5.06, P 0.05). A post hoc LSD test revealed that only in the 10-day exercise group DCX-immunoreactivity was significantly increased compared to sedentary controls (P 0.05). Regression analysis showed that both duration of running wheel access and distance ran during the last night signifi cantly contributed to the variation in DCX-expression (P and P 0.05, respectively). Hippocampal neurogenesis after normalization of activity levels Next, we studied whether the positive effect of exercise on hippocampal cell proliferation and the generation of new neurons persists or rapidly disappears when exercise is terminated. We measured the number of proliferating cells and immature neurons in the dentate gyrus of control mice, after 10 days of exercise, and after 10 days of exercise followed by 1 or 6 days of housing without a wheel. The experimental groups that had been housed with a wheel did not differ in the distance run on the last day (F(2,17)=1.00, n.s.). In agreement with the fi rst experiment, hippocampal cell proliferation was increased after the 10-day exercise period (Fig. 3A, one-way ANOVA: F(3,25)=5.87, P 0.01, post hoc LSD: P 0.05). However, cell proliferation had returned to baseline levels already 1 day after removal of the running wheel (P 0.05 compared to sedentary controls). In contrast, the number of DCXpositive immature neurons remained elevated for at least 6 days after the animals had lost 47

49 DCX-p os itive de ndr ite s Chapter 3 # K i- 67 po s it ive cells * *** *** sedentary 10d ex 1d dep 6d dep # * ** A B Figure 3: Hippocampal neurogenesis after removal of the running wheel. A) Ten days of wheel running signifi cantly increased the number of proliferating cells. However, when this 10-day exercise period was followed by 1 or 6 days of wheel deprivation, the number of proliferating cells immediately returned to baseline. B) Even 6 days after the running wheel had been removed from the home cage, the number of immature neurons remained signifi cantly increased. 10d ex: 10 days of exercise, 1d dep/6d dep: 1 day or 6 days of wheel deprivation after a 10-day period of exercise, respectively. #: P 0.10, *: P 0.05, **: P 0.01, ***: P Graphs show group averages ± S.E.M. 2 0 sedentary 10d ex 1d dep 6d dep access to the running wheel (Fig. 3B, one-way ANOVA: F(3,25)=3.20, P 0.05; post hoc LSD: P 0.05 after 1 and 6 days of wheel deprivation compared to sedentary controls). Discussion This study aimed to investigate the temporal dynamics of exercise-induced enhancement of hippocampal neurogenesis. Mice were housed with a running wheel for 1, 3 or 10 days and the number of Ki-67 and DCX-positive cells in the dentate gyrus were determined. Hippocampal cell proliferation gradually increased with prolonged running wheel exposure over a period from 1 to 10 days. This might suggest that a prolonged period of wheel access is required for exercise to have a stimulatory effect on hippocampal cell proliferation and neurogenesis. However, a more detailed analysis suggests that exercise may have an acute effect on cell proliferation when it has reached a certain threshold level. Important in this issue is that the distance animals ran per night was not constant, but gradually increased over 1, 3 and 10 days of wheel access. Regression analysis indicated that the distance run on the last day better explained the variation in cell proliferation than the number of days that the animals had access to the running wheel. Together, these data suggest that cell proliferation responds acutely to changes in physical exercise, but a certain minimal amount of activity is required. The idea that cell proliferation is directly responsive to physical exercise is sup- 48

50 Exercise and neurogenesis ported by the fi nding that hippocampal cell proliferation returned to baseline within a day after removal of the running wheel from the cage. Also, in agreement with this, we recently showed that the effects of exercise on cell proliferation were most prominent directly following the active dark phase, but had disappeared at the end of the resting phase, which also confi rms a direct effect of fluctuations in behavioral activity on the number of proliferating cells (Van der Borght et al., 2005a). The increase in cell proliferation after 3 and 10 days of exercise was reflected by an increase in the number of newly formed neurons after 10 days of wheel running. However, the reduction in Ki-67 positive cell number after removal of the running wheel was not followed by a reduction in the number of immature neurons. Six days after the running wheel had been removed from the home cage, the number of DCX-expressing cells remained elevated and even did not show a tendency towards a return to baseline levels. In the mouse, DCX is expressed in new hippocampal cells only for the fi rst 4 days after their formation (Kempermann et al., 2004). DCX-positive cells that we stained in the hippocampus of mice 6 days after removal of the running wheel therefore most likely represent cells that were generated after the period of increased physical activity had terminated. Since the generation of new cells returned to baseline levels immediately after removal of the wheel, our data indicate that the persistent increase in DCX immunoreactivity does not reflect an increase in the formation of new cells, but rather in the survival of cells that had been formed after running wheel activity had ended. Alternatively, our data could indicate that running persistently promoted differentiation towards a neuronal phenotype. However, this is a less likely explanation, since running has been reported to cause only a mild increase in the percentage of cells that acquires a neuronal phenotype (Van Praag et al., 1999b). There are numerous factors that are increased during physical activity and may underlie the acute effects of running on hippocampal cell proliferation. For instance, running leads to an increase in cerebral blood flow (Ide and Secher, 2000), which is accompanied by an increase in extracellular glucose and other nutrients that might be able to promote hippocampal cell production. The increase in extracellular glucose occurs rapidly after the onset of exercise and declines immediately when the period of enhanced activity is over (Bequet et al., 2001). The exercise-mediated increase in hippocampal neurogenesis might also be caused by increased release of certain neurotransmitters in the hippocampus. Hippocampal serotonin levels, for instance, were found to be elevated during physical exercise (Bequet et al., 2001; Gomez-Merino et al., 2001; Wilson and Marsden, 1996). Manipulations of the serotonergic system have been shown to affect hippocampal cell proliferation and the number of new neurons that is formed (Brezun and Daszuta, 2000; Malberg et al., 2000). Also, the medial septum becomes activated during voluntary movements, thereby inducing theta wave activity in the hippocampus (Teitelbaum et al., 1975). It has been shown previously that a lesion of the medial septum reduces hippocampal cell proliferation (Mohapel et al., 2005) and survival (van der Borght et al., 2005b), suggesting that an increased activity of the medial septum might stimulate neurogenesis. 49

51 Chapter 3 The positive effects of exercise on the generation of new neurons may also involve various growth factors. For instance, basic fibroblast growth factor (bfgf) mrna is signifi cantly upregulated during exercise (Gomez-Pinilla et al., 1997) and can stimulate hippocampal cell proliferation (Jin et al., 2003a). Also, peripheral blockade of vascular endothelial growth factor (vegf) completely abolishes the exercise-induced increase in neurogenesis. Levels of insulin-like growth factor I (IGF-1) in cerebrospinal fluid and brain are also known to be elevated during wheel running (Carro et al., 2000; Trejo et al., 2001). Furthermore, peripheral blockade of IGF-1 prevents the exercise-induced increase in nerogenesis (Trejo et al., 2001), indicating that IGF-1 may have an important role in the effects of exercise on the generation of new hippocampal neurons. Interestingly, vegf and IGF-1 are both factors involved in angiogenesis (Lopez- Lopez et al., 2004). Hippocampal cell proliferation has been reported to be closely associated with the vasculature (Palmer et al., 2000). In vitro studies have also shown that coculturing of endothelial cells with neural stem cells stimulated self-renewal of the stem cells and inhibited differentiation (Shen et al., 2004). Therefore, enhanced neovascularization might positively influence progenitor proliferation or survival of newly formed neurons in the hippocampus. IGF-1, and also BDNF, might also be involved in the exercise-mediated increase in survival of newly formed cells by activating the PI3K/Akt signaling cascade, which is involved in neuronal survival (Chen and Russo-Neustadt, 2005; Feldman et al., 1997). IGF- 1 is involved in granule cell survival in the postnatal dentate gyrus (Cheng et al., 2001) and it has been shown to be neuroprotective in cerebral ischemia models (Guan et al., 1993). BDNF is also strongly increased in the hippocampus by physical exercise (Adlard et al., 2004; Neeper et al., 1996) and has well-known neuroprotective effects, especially on newly formed neurons (Lee et al., 2002; Sairanen et al., 2005). However, the temporal dynamics of the increase in IGF-1 and BDNF expression during exercise and after normalization of activity levels remain to be investigated. In summary, we showed that hippocampal cell proliferation is acutely regulated by physical activity, whereas the exercise-induced increase in newly formed cell survival persisted for at least 6 days after removal of the running wheel from the cage. These fi ndings suggest that exercise has effects on cell proliferation and cell differentiation or cell survival that are mediated by at least partly different mechanisms. Additional experiments are required to establish what these mechanisms are. 50

52 4 Chapter Input from the medial septum regulates adult hippocampal neurogenesis Karin Van der Borght 1, Jan Mulder 1, Jan N. Keijser 1, Bart J. L. Eggen 2, Paul G. M. Luiten 1, Eddy A. Van der Zee 1 1) Department of Molecular Neurobiology, Graduate school of Behavioural and Cognitive Neurosciences, and 2) Department of Developmental Genetics, Groningen Biomolecular Sciences and Biotechnology Institute; University of Groningen, P. O. Box 14, 9750 AA, Haren, The Netherlands Brain research Bulletin (2005), 67(1-2):

53 Chapter 4 Abstract Neural progenitors in the subgranular zone of the hippocampal formation form a continuously proliferating cell population, generating new granule neurons throughout adult life. Between ten days and one month after their formation, many of the newly generated cells die. The present study investigated whether a partial lesion of one of the main nuclei projecting to the hippocampus, the medial septum (MS), aff ects survival and diff erentiation of cells during this critical period. Rats were injected with BrdU and fi ve days later excitotoxic lesion of the MS was applied by infusion of either 30 nmol or 60 nmol of N-methyl-D-aspartate (NMDA). One week after the lesion, quantifi cation of immunopositive cells revealed that the number of GABAergic cells was signifi cantly reduced in both lesioned groups, whereas a decline in cholinergic cell number was observed only after injection of 60 nmol of NMDA. The partial septohippocampal denervation signifi cantly reduced hippocampal neurogenesis. Survival of newly generated neurons was decreased by approximately 40%. The MS lesion did not aff ect proliferation of hippocampal progenitors. The present study indicates the importance of a functional septohippocampal pathway for the regulation of hippocampal neurogenesis and it highlights the potential role of GABA as a mediator in this phenomenon. 52

54 The medial septum regulates neurogenesis Introduction Throughout adult life, neural progenitors in the subgranular zone (SGZ) of the dentate gyrus (DG) give rise to new neurons in the granule cell layer (GCL) (Altman and Das, 1965; Kaplan and Hinds, 1977; Stanfield and Trice, 1988). Adult hippocampal neurogenesis has been described in several species, including rodents, non-human primates and humans (Eriksson et al., 1998; Gould et al., 1997; Kornack and Rakic, 1999). Double labeling of the thymidine analogue 5 -bromo-2 -deoxyuridine (BrdU) and the proliferation marker Ki-67 at different time points after injection with BrdU showed that cell progenitors remain proliferating for approximately four days (Dayer et al., 2003). After the cells become postmitotic, they differentiate into granule cells, send axons to the CA3 region (Hastings and Gould, 1999) and receive input from other cells (Markakis and Gage, 1999). About 50% of the newly generated neurons die between ten days and one month after their birth (Dayer et al., 2003; Hastings and Gould, 1999). Interestingly, there is evidence that many of the newly formed cells can be rescued by training animals in a hippocampus-dependent learning task or housing them under enriched conditions (Ambrogini et al., 2000; Gould et al., 1999a; Kempermann et al., 1997). However, it is still unknown which factors determine if a newly formed cell will survive and be integrated in the hippocampal circuitry. The present study was aimed at investigating whether the input from the medial septum (MS) into the hippocampus is involved in the regulation of newly formed cell survival and differentiation. The MS consists of cholinergic neurons as well as gammaaminobutyric acid (GABA)-producing interneurons and projection neurons. The cholinergic neurons innervate the three major cell types of the hippocampus, i.e. the pyramidal, non-pyramidal and granule cells (Frotscher, 1989; Frotscher and Misgeld, 1989; Gaykema et al., 1990), whereas GABAergic projection neurons mainly innervate hippocampal interneurons (Freund and Antal, 1988; Gulyas et al., 1991). The septohippocampal pathway has been shown to be pivotal for proper hippocampal functioning. Memory performance can be correlated with cholinergic activity in the hippocampus, measured by high-affi nity choline uptake or choline acetyltransferase (ChAT) activity (Decker et al., 1988; Dunbar et al., 1993). Moreover, mechanical disruption of the septal projections to the hippocampus (Alvarez-Pelaez, 1973; Olton, 1977) or lesions of the MS (Hagan et al., 1988; Johnson et al., 2002) disturb memory retention. Considering the key role of the basal forebrain in cognitive functions, the learning-induced increase of adult neurogenesis (Ambrogini et al., 2000; Gould et al., 1999a) could potentially be caused by an increased activation of the septohippocampal pathway. Moreover, during aging a loss in MS cholinergic and GABAergic cells has been reported, which leads to an impairment in hippocampal functioning (Apartis et al., 2000; Fischer et al., 1989; Fischer et al., 1992; Krzywkowski et al., 1995). Concomitantly with the decrease of cell number in the MS, hippocampal neurogenesis significantly declines during aging (Heine et al., 2004a; Kuhn et al., 1996). Additional data also suggest a role for the MS in the regulation of adult hippocampal neurogenesis. Wheel running, for instance, which 53

55 Chapter 4 has been shown to evoke hippocampal theta waves, induced by synchronized signaling of septal cholinergic and GABAergic neurons (Teitelbaum et al., 1975), robustly increases hippocampal neurogenesis in mice and rats (Kim et al., 2003; Van Praag et al., 1999b). Moreover, a recent study by Cooper-Kuhn et al. (2004) has shown that 192IgG-saporin infusion into the lateral ventricle, causing cholinergic lesion of the basal forebrain including the MS, reduces hippocampal neurogenesis. In the present study, rats were injected with BrdU in order to gain a deeper insight into the role of MS-derived cholinergic and GABAergic projections to the hippocampus on the regulation of hippocampal neurogenesis, Five days after the last BrdU injection, N-methyl-D-aspartate (NMDA) was infused into the MS to induce an excitotoxic lesion. One week after the lesion, the impact of a reduced septohippocampal innervation on hippocampal neurogenesis was investigated using immunocytochemistry for BrdU and the proliferation marker Ki-67. Materials and Methods Animals and experimental procedure Adult male Wistar rats ( g, n=21), bred in our own facilities, were used. Rats were housed individually, had free access to water and food and were kept under a 12/12h light/ dark cycle (lights on at 08.00h) in a temperature-controlled environment (21±2 C). Animals were injected intraperitoneally with 100 mg/kg of BrdU (Sigma, St. Louis, MO, USA) dissolved in saline (20 mg/ml, ph 7.0), daily for three consecutive days. Seven days after the fi rst BrdU injection, animals were deeply anesthetized with 2%-2.5% isoflurane and mounted in a stereotaxic frame (Narishige, Japan). Using a Hamilton microsyringe (Bonaduz, Switzerland), 1 μl of 0.01 M phosphate-buffered saline (PBS) containing 30 nmol (n=7) or 60 nmol (n=7) of NMDA (Sigma) was slowly injected into the MS (coordinates from Bregma: AP: 0.2 mm, ML: 0.0 mm DV: 6.3 mm and 6.0 mm (Paxinos and Watson, 1986)) at an angle of 5 with the vertical plane, in order to avoid the sagittal sinus. Control animals (n=7) were injected with 0.01 M PBS. One week after surgery, animals were sacrificed under deep anesthesia with 1 ml of sodium-pentobarbital by transcardial perfusion with saline, followed by a fi xative solution consisting of 2.5% paraformaldehyde and 0.05% glutaraldehyde in 0.1 M phosphate buffer (ph 7.4). Prior to perfusion, after the thorax had been opened, a blood sample was taken from the heart. Blood samples (approximately 0.5 ml) were collected in prechilled Eppendorf tubes containing EDTA as anticoagulant. After centrifugation at 2600 rpm for 15 min, the plasma was collected, and stored at -80 C for later radioimmunoassay analysis of corticosterone (ICN Biomedicals, Costa Mesa, USA). Brains were placed in 0.01 M PBS overnight and subsequently cryoptrotected in 30% sucrose at 4 C. Three series of coronal sections (30 μm-thick) spanning the MS (Bregma 1.20 to 0.26) and fi fteen series through the entire extent of the hippocampus (Bregma 2.12 to 6.30) were cut on a cryostat and collected in 0.01 M PBS containing 54

56 The medial septum regulates neurogenesis 0.1% sodium azide. All procedures concerning animal care and treatment were in accordance with the regulations of the ethical committee for the use of experimental animals of the University of Groningen (DEC number 2729) and European Community Council Directives. Immunocytochemistry All immunocytochemical procedures were performed on free-floating sections. Immunocytochemistry for glial fibrillary acidic protein (GFAP), glutamic acid decarboxylase (GAD) 65/67, Ki-67 and ChAT was performed following a similar protocol. In brief, sections were pretreated with 0.3% H 2 O 2 for 30 min. Nonspecific binding of immunoreagents was blocked with 3% normal goat serum (Zymed, San Francisco, CA, USA). Subsequently, sections were incubated overnight at 4 C with mouse monoclonal anti-gfap (1:1000) or rabbit polyclonal anti-gad65/67 (1:5000; both purchased from Sigma, St Louis, MO, USA), for 48 h with mouse monoclonal anti-ki-67 (1:200; Novocastra, Newcastle upon Tyne, UK) or for 72 h with goat polyclonal anti-chat (1:200; Chemicon, Harrow, UK). After a second blocking step, the biotinylated secondary antibodies (goat-anti-mouse, goatanti-rabbit and rabbit-anti-goat IgGs, all 1:400; Jackson Immunolabs, West Grove, PA, USA) were added. This was followed by incubation with avidin-biotin-complex (1:400; ABC Elite kit, Vector Laboratories, Burlingame, CA, USA). Cells were visualized using diaminobenzidine (DAB) as chromogen (20 mg/100 ml). BrdU immunocytochemistry started with DNA denaturation procedures (van der Borght et al., 2005a). In brief, sections were incubated for 2 h at 65 C in 2 x saline sodium citrate (2xSSC) containing 50% formamide. After rinses with 2xSSC, 2 M HCl (37 C for 30 min) and 0.1 M borate buffer (ph 8.5), sections were exposed to the primary antibody (rat monoclonal anti-brdu, 1:800, Oxford Biotechnology, Oxfordshire, UK) overnight at 4 C. As a secondary antibody biotinylated donkey-anti-rat IgGs (1:400; Jackson Immunolabs) were used. The staining was developed using DAB and H 2 O 2. For triple labeling for BrdU, GFAP and the neuronal marker NeuN, a similar DNA denaturing procedure as described above was used. Subsequently, sections were incubated for 72 h in the primary antibody solution containing rat monoclonal anti-brdu (1:200, Oxford Biotechnologies), rabbit polyclonal anti-cow GFAP (1:400; DAKO, Glostrup, Denmark) and mouse monoclonal anti-neun (1:400; Chemicon, Temecula, CA, USA). The following secondary antibodies were used: biotinylated donkey-anti-rat, Cy5- conjugated donkey-anti-rabbit and rhodamine red-conjugated donkey-anti-mouse F(ab ) fragments (1:200; all from Jackson Immunolabs). BrdU staining was made visible by incubation with Fluorescein (DTAF)-conjugated streptavidin (1:200; Jackson Immunolabs). Data analysis The material was examined in bright-fi eld illumination and confocal microscopy (Zeiss LSM510 confocal laser and scanning microscope) was used for multiple immunofluorescence. NMDA infusion into the medial septum results in widespread neuronal damage, often exceeding the target nucleus (Harkany et al., 2000). Following injury, astrocytes start to proliferate and become activated, a process which is called reactive gliosis. 55

57 Chapter 4 Therefore, the presence of dense GFAP-immunoreactivity can be considered as a suitable indicator of neuronal damage (Ridet et al., 1997). In the intact animal, hardly any GFAPpositive astrocytes in the MS are present. After infusion of NMDA, a very dense cluster of GFAP-positive cells is visible at the injection site. This area we called the core of the lesion. The less-dense GFAP-positive rim surrounding the core was considered to be the penumbra zone (Horvath et al., 2002). Guided by the GFAP staining, animals were selected in which the area with the highest density of GFAP-positive cells was located in the MS and only those animals were included in further analyses (30 nmol NMDA: n=6, 60 nmol of NMDA; n=5). Quantitative analyses were based on cell counts, which were performed blindly as to the treatment of the animals. For the quantification of the number of ChAT and GAD65/67-positive cells in the MS, 4-6 representative sections of each animal, ranging from Bregma 1.20 to 0.26 and equally spaced and at matched anteroposterior levels, were used. All immunopositive cells in the MS were counted with a 20x objective. Cells were counted throughout the entire z-axis of every section. To avoid the inclusion of fragmented cell profi les, only cells with clear cytoplasmic staining and discernible, unstained nuclei were counted. Only immunopositive cells in the MS, and not in the vertical limb of the diagonal band of Broca, were included. Whenever the border between the MS and the vertical limb of the diagonal band was unclear, only those cells were counted that were located dorsally from the anterior commissure. In addition, the surface area of the MS was measured using a computer-based image analysis system (Quantimet W500, Leica, Rijswijk, The Netherlands). By calculating the number of counted cells/mm2, we obtained a standardized measure for the average cell density in the MS of each animal. For the quantification of the number of Ki-67- and BrdU-positive cells in the GCL and the SGZ of the DG every 15th section containing the hippocampus was taken, revealing a total of 12 sections per animal throughout its anteroposterior extent. Quantifi - cations were performed using a 40x objective. Only Ki-67 and BrdU immunostained cell nuclei, throughout the entire thickness (30 μm) of the section, on the border of the hilus and the GCL were counted, including cells that were located one cell diameter deviating from this border. BrdU-positive cells laying in the GCL were counted as well, since newly formed cells that are 12 to 14 days of age may have migrated into the GCL. The total number of counted cells was multiplied by fi fteen to obtain an estimation of the total number of positive cells per DG (Gould et al., 1999a; Malberg et al., 2000). A multi-track analysis of the BrdU/GFAP/NeuN triple staining was performed using a Zeiss LSM510 confocal laser and scanning microscope. From the three experimental groups (n=3 per group), fi fty BrdU-positive cells were randomly chosen and scanned in its entire z-axis, using 1 mm intervals, in order to exclude false double labeling due to an overlay of signals from different cells. With the help of specialized software (Zeiss LSM Image Browser Version ), each scanned BrdU-positive cell was qualified to be immunoreactive for GFAP or NeuN. 56

58 The medial septum regulates neurogenesis MS 30 nmol 60 nmol Figure 1: Reconstruction of the lesions. The grey area illustrates the size and location of a representative lesion. The penumbra zones of all lesions were located within the area delineated by the black line. The core of the lesion was found between Bregma 0.48 and 0.70, whereas the penumbra ranged from Bregma 1.20 to The anteroposterior extent did not diff er between the two concentrations of NMDA. Statistics Statistical analyses were performed to test potential differences in group average between the three experimental groups, for the following parameters: 1) ChAT-positive cell numbers and GAD65/67-positive cell numbers in the medial septum, 2) BrdU and Ki-67 positive cell counts in the dentate gyrus, 3) the distribution of the different phenotypes within the BrdU-positive cell population and 4) the plasma corticosterone concentrations. All these parameters were analyzed using one-way analysis of variance (ANOVA). Whenever this revealed significant differences, pairwise comparison was performed using the post-hoc Tukey-HSD test. All data are expressed as means ± standard error of the mean (S.E.M.). 57

59 Chapter 4 Results Lesion The size and location of the lesions were determined on the basis of the GFAP immunostaining. A lesion was considered to be correctly placed when the core and/or the penumbra of the lesion overlapped with the MS. Based on this criterion, three animals (one of the 30 nmol group and two of the 60 nmol group) were excluded from further analyses, which resulted in a fi nal group size of 7 sham animals, 6 rats with a lesion of 30 nmol of NMDA and 5 rats with a 60 nmol lesion. A representative lesion, as well as the delimitation of the area included in the other lesions are presented in Fig. 1. In order to determine the extent of the damage to the cholinergic system caused by the lesion, an immunocytochemical staining for the acetylcholine-synthesizing enzyme ChAT was performed (Fig. 2A-C). The lesion with 60 nmol of NMDA resulted in a 32% and significant reduction of the number of ChAT-positive neurons in the MS (F(2,17)=7.48, P<0.05; post hoc P<0.05). Thirty nmol of NMDA did not affect the cholinergic cell number in the MS. Lesion-induced changes in the GABAergic system were visualized by immunoreactivity for the 65-kDa and the 67-kDa isoforms of GAD, the enzyme that converts glutamic acid into GABA (Fig. 2D-F). A significant difference was found between the three groups (F(2,17)=11.93, P<0.001). Post-hoc testing indicated a 62% and significant reduction in GAD65/67- immunopositive cells in rats that were lesioned with 30 nmol of NMDA (P<0.01). The 60 nmol concentration caused a 66% and signifi cant reduction of GAD65/67-positive neurons compared to controls (P<0.01). These results demonstrate a potent negative effect of NMDA-infusion on the number of GABAergic cells in the MS and a moderate reduction in cholinergic cell number in the MS only after infusion with 60 nmol of NMDA. Neurogenesis In order to investigate whether a reduced input from the MS to the hippocampus had an effect on the survival of newly formed cells, all animals were injected with BrdU one week before surgery and the number of BrdU-positive cells still present in the GCL one week after the lesion was quantified (Fig. 3A,B, page 140). BrdU immunocytochemistry demonstrated that NMDA infusion into the MS considerably and significantly reduced the survival of newly formed cells in the GCL (F(2,17)=12. 82, P=0.001). The lesion with 30 nmol of NMDA resulted in a 39% and significant decline of the number of BrdU-positive cells (P<0.01), whereas infusion of 60 nmol of NMDA caused a decrease of 40% (P<0.001) suggesting an important role for the MS in the regulation of newly formed cell survival. To determine whether the MS lesion specifically affected the survival of newly formed astrocytes or neurons, a triple labeling for BrdU, the glial marker GFAP and the neuronal marker NeuN was performed (Fig. 3C,D, page 140). Analysis revealed that the phenotypes of the BrdU-positive cells did not differ between the 30 nmol and the 60 nmol 58

60 Number of Ch A T-positive cells/mm 2 The medial septum regulates neurogenesis A 100 * sham 30 nmol 60 nmol Number of GAD65/67-positive cells/mm ** ** sham 30 nmol 60 nmol B Sham C 60 nmol E Sham F 30 nmol D Figure 2: Impact of NMDA infusion into the MS on the cholinergic and GABAergic cell number A) Thirty nmol of NMDA did not aff ect the number of ChAT-positive cells in the MS, whereas 60 nmol of NMDA caused a signifi cant reduction (* P<0.05, post-hoc analysis following ANOVA). Representative photomicrographs of ChAT-immunostaining are shown in panels B (sham) and C (60 nmol of NMDA). Scale bar=80 mm. D) Both concentrations of NMDA caused a signifi cant decline in the number of GAD65/67-positive cells (** P<0.01, posthoc analysis following ANOVA). Photomicrographs of GAD65/67-positive cells in the MS are shown in E (sham) and F (30 nmol of NMDA). Scale bar=100 μm. groups, and the data from the two lesioned groups were therefore pooled. Z-series of randomly chosen BrdU-positive cells throughout the entire GCL demonstrated that in sham animals as well as in lesioned rats, the percentage of BrdU-positive cells colocalizing with either NeuN or GFAP was similar. This result indicates that the MS lesion reduced the number of all newly formed cells, regardless of their phenotype. Mitotic cells express the nuclear marker Ki-67 throughout all phases of the cell cycle (Kee et al., 2002; Scholzen and Gerdes, 2000). Therefore, staining of Ki-67 positive cells provides information about the number of cells that were in any phase of the cell cycle (except for G0) at the time of termination. In this part of the study, immunocytochemical staining for Ki-67 was used to determine whether a decrease in the septohippocampal input had an effect on the number of proliferating cells in the SGZ. No differences were detected between the lesioned groups and the sham-operated animals, though the two lesioned groups differed from each other (F(2,17)= 4.34, P<0.05; post hoc: P<0.05). The 59

61 Chapter 4 lesion with 30 nmol of NMDA resulted in 57% more Ki-67 positive cells in the SGZ compared to the lesion with 60 nmol of NMDA (Fig. 3E,F, page 140). Since MS lesions occasionally increase plasma corticosterone levels (Alema et al., 1995), which can negatively affect hippocampal neurogenesis (Gould et al., 1997; Heine et al., 2004b; Pham et al., 2003), blood samples were taken prior to perfusion to determine plasma corticosterone concentrations. No changes were detected (sham: 19.6 ± 4.6 μg/dl, 30 nmol of NMDA: 23.0 ± 6.1 μg/dl, 60 nmol of NMDA: 20.9 ± 4.1 μg/dl), indicating that the observed changes in neurogenesis were not likely to be caused by changes in corticosteroid concentrations. Discussion The present study investigated whether a partial lesion of the MS, a region of the basal forebrain that densely innervates the hippocampal formation, has an effect on neurogenesis in the hippocampus of the adult rat. In order to investigate this, the MS was damaged by infusion of two different concentrations of NMDA. Both 30 nmol and 60 nmol of NMDA significantly reduced the survival of cells that had been generated in the GCL fi ve to seven days prior to the lesion. This reduced survival was observed both for new neurons and newly formed astrocytes. Excitotoxic lesion of the MS did not affect cell proliferation in the SGZ, one week after the lesion. Our data show that the lesion of the MS with 30 nmol and 60 nmol of NMDA caused a similar reduction in the number of GABAergic cells. The number of cholinergic cells was reduced only after infusion of 60 nmol of NMDA into the MS. The decrease in hippocampal neurogenesis was comparable between the two lesioned groups, which suggests that the GABAergic system, and not the cholinergic input to the hippocampus, is involved in the regulation of survival of newly generated hippocampal granule neurons. The GABAergic system may affect survival of newly formed cells either directly or indirectly, via synaptic or nonsynaptic signaling. The effects of the lesion on survival of newly generated cells may have been caused by nonsynaptic effects of GABA. Nonsynaptic GABAergic signaling takes place during various steps of nervous system development (Owens and Kriegstein, 2002), such as synaptogenesis (Belhage et al., 1998), neuronal differentiation (Nguyen et al., 2003) and neurite extension (Wolff et al., 1978). It may therefore be possible that GABA also acts as a neurotrophic factor for immature or nearly mature hippocampal neurons that are formed during adulthood. Because infusion of NMDA into the MS strongly decreased the GABA-ergic input to the hippocampus, it could be hypothesized that this caused a reduction of neurotrophic support for the survival of newly generated cells in the DG. Alternatively, the effects of the lesion on newly formed cell survival may be the result of changes in synaptic signaling. It has been shown that newly generated dentate granule neurons receive synaptic GABAergic input. Studies with acute hippocampal slices 60

62 The medial septum regulates neurogenesis demonstrate that newly generated granule neurons exhibit postsynaptic responses that are typical for fast neurotransmitters, such as GABA (Van Praag et al., 2002). In intact conditions, GABAergic projection neurons in the MS innervate GABAergic interneurons in the hippocampus. The GABAergic innervation of hippocampal interneurons is believed to lead to a disinhibition of hippocampal principal cells (Freund and Antal, 1988; Krnjevic et al., 1988; Toth et al., 1997). Following this line of reasoning, a loss of septal GABAergic projection neurons may therefore result in increased inhibition of the hippocampus. This is also suggested by other studies, based on the fi nding that a selective lesion of GABAergic septal neurons prevented the induction of theta wave activity in the hippocampus (Wu et al., 2002; Yoder and Pang, 2005). The assumption of disturbed disinhibition supports the hypothesis that certain forms of increased hippocampal activity, caused, for instance, by enriched housing conditions or trace eyeblink conditioning, promote the survival of newly generated neurons (Gonzalez-Lima et al., 1994; Gould et al., 1999a; Kempermann et al., 2002; Puurunen et al., 2001). The almost exclusive decrease of GABAergic cells cannot rule out the possibility that other factors besides GABA may have accounted for the observed effects on the survival of newly formed cells. Neurons of the basal forebrain also express certain peptides that are anterogradely transported to the hippocampus and have been reported to stimulate hippocampal neurogenesis (Lai et al., 2003; Machold et al., 2003). In addition, the cholinergic and the GABAergic system are closely interconnected (Brauer et al., 1998; Van der Zee and Luiten, 1994; Wu et al., 2000). The net output from the MS to the hippocampus depends on the balance between cholinergic and GABAergic activity in the MS, indicating that the loss of GABAergic cells in the MS after the lesion may also have affected the properties of the cholinergic septal neurons. The proliferation of hippocampal progenitors was not influenced by the lesion. In vitro studies indicate a role for acetylcholine in cell progenitor proliferation. Cholinergic stimulation enhances proliferation of embryonic cortical neural precursors (Li et al., 2001; Ma et al., 2000; Zhao et al., 2003) and of oligodendrocyte progenitors (Larocca and Almazan, 1997). In addition, Lai et al. (2003) demonstrated that disruption of the septohippocampal connection by a lesion of the fi mbria-fornix results in a reduced proliferation in the adult mouse DG. Recent studies show that extensive lesion of cholinergic cells in the basal forebrain, by infusion of 192-IgG saporin, strongly reduce cell proliferation in the subventricular zone and the olfactory bulb (Calza et al., 2003) and also leads to a moderate, but significant, reduction in hippocampal cell proliferation (Mohapel et al., 2005). In the present study, the effect of the MS lesion on hippocampal cell proliferation was less obvious. Although progenitor proliferation in the two lesioned groups differed from each other, the results did not deviate from sham animals. This may be due to the fact that even the highest concentration of NMDA only moderately reduced cholinergic cell number in the MS, which may not be sufficient to affect cell proliferation. The observed decline in hippocampal neurogenesis after the MS lesion may reflect the changes that occur during aging. The MS of aged rats contains less cholinergic and GA- BAergic cells than in younger animals (Fischer et al., 1989; Krzywkowski et al., 1995) and 61

63 Chapter 4 the reduced cell number correlates with impairments in hippocampus-dependent learning (Fischer et al., 1989; Fischer et al., 1992). It has also repeatedly been reported that hippocampal neurogenesis is decreased during aging (Heine et al., 2004a; Kempermann et al., 2002; Kuhn et al., 1996). Twenty-one-month-old rats, for instance, show a 90% reduction in cell proliferation, when compared with 6-month-old animals (Kuhn et al., 1996). However, the survival of newly formed hippocampal neurons is not affected during aging (Bondolfi et al., 2004; Heine et al., 2004a). Since we show here that the GABAergic component of the septohippocampal pathway is unlikely to induce changes in hippocampal cell proliferation, but specifi cally seems to affects cell survival, the reduction in cell proliferation occurring during aging is probably not caused by the reduction in GABAergic cell number in the MS. In summary, here we show that NMDA infusion into the MS resulted in a decline in neurogenesis in the DG. This is caused by a reduction in the number of surviving newly formed granule neurons and not in the number of proliferating progenitors. Our data suggest that the GABAergic system may be one of the factors that determine the fate of newly generated granule neurons. These results provide new insight into the possible mechanisms underlying the regulation of survival of hippocampal granule neurons generated during adulthood. 62

64 5 Chapter Effects of active shock avoidance learning on hippocampal neurogenesis and plasma levels of corticosterone Karin Van der Borght 1, Peter Meerlo 1, Paul G.M. Luiten 1, Bart J.L. Eggen 2, Eddy A. Van der Zee 1 1) Department of Molecular Neurobiology, Graduate school of Behavioural and Cognitive Neurosciences, and 2) Department of Developmental Genetics, Groningen Biomolecular Sciences and Biotechnology Institute; University of Groningen, P. O. Box 14, 9750 AA, Haren, The Netherlands Behavioural Brain Research (2005), 157(1):

65 Chapter 5 Abstract Hippocampal granule neurons that are newly formed during adulthood might be involved in learning and memory processes. Experimental data suggest that only hippocampus-dependent learning tasks stimulate neurogenesis. To further address this issue, the eff ects of active shock avoidance (ASA) learning on hippocampal progenitor proliferation and survival of newly formed cells were investigated. ASA training, although considered as hippocampus-independent, is known to induce several neurobiological alterations in the hippocampus. Adult Wistar rats were trained in a shuttle box using a one-day or four-day paradigm and brains were analyzed for the mitotic marker Ki-67. Eff ects on survival of newly generated cells were examined by immunocytochemistry for 5-Bromo-2-deoxyuridine (BrdU), which was injected one week before the training. Neither proliferation nor survival was aff ected by the ASA learning task. Because elevated glucocorticoid levels have a negative impact on hippocampal neurogenesis, blood samples were taken throughout the four-day training paradigm. Both trained animals and control rats that were only placed in the shuttle box without receiving foot shocks showed a similar rise in corticosterone, enabling us to exclusively investigate the eff ects of ASA learning on neurogenesis without diff erential interference of stress between groups. On the other hand, the fi nding that ASA induced elevations in plasma corticosterone, but did not infl uence proliferation or survival of newly formed cells, indicates that this type of stress does not aff ect neurogenesis. The present study shows that, in line with the existing data on other hippocampus-independent learning tasks, ASA training has no eff ect on hippocampal neurogenesis. 64

66 Active shock avoidance learning and neurogenesis Introduction The dentate gyrus (DG) of the hippocampal formation remains capable of generating new neurons during adulthood. The formation of these new cells depends on rapidly dividing progenitors residing in the subgranular zone (SGZ). Upon migration into the hippocampal granule cell layer (GCL), these cells differentiate into a neuronal phenotype (Cameron et al., 1993; Okano et al., 1993), form mossy fi bers (Hastings and Gould, 1999), receive input from other cells (Markakis and Gage, 1999) and ultimately become functional granule neurons (Van Praag et al., 2002). Although adult neurogenesis has received much attention during the last decade, the function of newly generated hippocampal neurons is still unclear. Since the hippocampus plays a key role in learning and memory (reviewed by (Holscher, 2003; Knierim, 2003)), it has been hypothesized that newly formed neurons are involved in one or more aspects of learning (Gould et al., 1999c). This hypothesis is supported by the observation that factors known to facilitate learning performance, such as estrogens (Horvath et al., 2002; Sherwin, 1988), running wheel activity (Samorajski et al., 1985; Van Praag et al., 1999a) or environmental enrichment (Cooper and Zubek, 1958; Frick et al., 2003), also have a stimulatory effect on adult neurogenesis (Kempermann et al., 1997; Tanapat et al., 1999; Van Praag et al., 1999a). Aging, on the other hand, or high glucocorticoid levels, which can have a negative impact on learning (Bowman et al., 2003; McNamara et al., 1977; Newcomer et al., 1994), reduce hippocampal neurogenesis (Cameron and Gould, 1994; Kuhn et al., 1996). Several studies have been performed to investigate the connection between learning and neurogenesis. Experimental data on progenitor proliferation or survival of newly formed neurons exist for Morris water maze (MWM) learning and trace eyeblink conditioning (TEC), but the results are still rather controversial (Dobrossy et al., 2003; Drapeau et al., 2003; Gould et al., 1999a; Lemaire et al., 2000; Merrill et al., 2003; Van Praag et al., 1999a), possibly caused by differences in species, gender and timing of BrdU injections (Greenough et al., 1999). In view of the fact that so far only the hippocampus-dependent and not the hippocampus-independent versions of MWM learning and TEC have been found to be able to induce changes in hippocampal neurogenesis, it has been postulated that only learning tasks that are dependent on the hippocampus have the capability to affect neurogenesis (Gould et al., 1999a). Surprisingly, most of the work concerning learning and neurogenesis has been limited to only two specific learning tasks, MWM and TEC. Because of the differential results that were obtained using these two learning tasks, it is interesting to investigate whether an essentially different type of learning, such as active shock avoidance (ASA), is able to induce changes in hippocampal neurogenesis. ASA is a classical Pavlovian conditioning task that involves non-declarative memory. Acquisition of this task is not supposed to be strictly dependent on the hippocampus, since there is an overlap between the conditioned and the unconditioned stimulus, meaning that no temporal gap between both stimuli has 65

67 Chapter 5 # Active avoidances # Active avoidances responders non-responders responders non-responders Block of 5 trials Block of 5 trials A B Figure 1: ASA learning curves. Data are presented as the number of active avoidances per block of 5 trials. Responders are rats that reached the learning criterion of four successive active avoidances: 9 out of 18 animals in the one-day ASA protocol (A) and 6 out of 10 rats in the four-day protocol (B). to be bridged by the hippocampus (Solomon et al., 1986; Wallenstein et al., 1998; Weiss et al., 1999). If the hypothesis was correct that only hippocampus-dependent learning tasks can influence neurogenesis, ASA acquisition should not have an effect on hippocampal neurogenesis. Yet, multiple studies have demonstrated that the hippocampal formation, although not necessary for acquisition of ASA, is certainly affected by this task (Ramirez and Carrer, 1989; Van der Zee and Luiten, 1999; Van Reempts et al., 1992), which emphasizes the importance of investigating the effects of ASA training on hippocampal neurogenesis. In the present study, rats were trained in the ASA task during one or four days. Two different training protocols were used, because we aimed to investigate whether a short one-day protocol had a different effect on hippocampal neurogenesis than a more prolonged training of four days, which requires repeated activation of the hippocampus. Twenty-four hours after the last training session, animals were sacrifi ced and brains were analyzed for progenitor proliferation in the SGZ with Ki-67 immunocytochemistry and for survival of newly formed granule cells, which was achieved by staining for the thymidine analogue BrdU that had been injected one week before training. In addition, since ASA is supposed to be a stressful learning task and because several studies have reported that stress has a negative impact on adult neurogenesis (Czeh et al., 2001; Falconer and Galea, 2003; Gould et al., 1997; Pham et al., 2003; Tanapat et al., 2001), blood samples were taken during the experiment to determine plasma levels of corticosterone (CORT). 66

68 Active shock avoidance learning and neurogenesis Materials and Methods Animals and housing Fifty-seven male Wistar rats (circa 300 g, bred in our own facility) were housed individually, had free access to food and water and were kept in climate rooms with a 12/12h lightdark cycle (lights on at 8:00 a.m.). All procedures concerning animal care and treatment were in accordance with the regulations of the ethical committee for the use of experimental animals of the University of Groningen (DEC 2719). ASA training In both experiments, ASA testing was performed during the light phase of the animals, using a fully automated shuttle box (Coulbourn Inst. L.L.C., Allentown, USA), which consisted of two identical compartments (25x25x30 cm), separated by a low Perspex hurdle. Scrambled foot shocks were delivered through the grid floor. Whenever an animal moved from one compartment to the other, this was detected by infrared emitters and receivers. In both training paradigms a random intertrial interval of 20, 40 or 60 s was used. After a habituation period of 3 minutes, the conditioned stimulus (CS, a 4.5 khz tone) was presented. Five seconds after the onset of the CS, the animals received a mild foot shock of 0.3 ma for maximally 3 s, which served as the unconditioned stimulus (US). If the rat failed to make a response, both CS and US were terminated after 3 s of foot shock. If the animal jumped to the other compartment during CS presentation, this was recorded as an active avoidance. Whenever it jumped to the other side when both CS and US were presented, this was recorded as an escape response. CS and US were terminated immediately if the animal moved to the other compartment. Animals for the one-day training paradigm (n=26) were intraperitoneally injected with 50 mg/kg 5-Bromo-2-deoxyuridine (BrdU, Sigma, St. Louis, USA), dissolved in saline (20 mg/ml) on 3 consecutive days. Seven days after the fi rst injection, part of the animals (n=18) were trained in the one-day ASA test, which consisted of 40 trials. The rest of the group (n=8) served as habituated controls and were placed in the shuttle box for 30 min without exposure to CS and US. Animals for the four-day training paradigm (n=30) received three injections on consecutive days with 100 mg/kg BrdU, in order to label more newly generated cells (Cameron and McKay, 2001). Seven days after the fi rst injection, ten animals were trained, using 15 trials per day for four days. There were two control groups: habituated controls (n=10) that were placed in the shuttle box for 10 min per day and a group of home cage controls (n=10). Blood sampling Throughout the four-day experiment, blood samples were taken and plasma CORT levels were measured with a radio-immuno assay (ICN Pharmaceuticals, Costa Mesa, USA). Sampling took place via the tail bleeding method (Meerlo et al., 2002). On all four training days and the day after training, baseline blood samples (200 μl) were taken between 9:00 67

69 Chapter 5 a.m. and 10:00 a.m. from trained animals and shuttle box control animals. From the home cage control animals only three baseline samples were taken throughout the experiment. Twenty min after the start of the training or habituation, a second daily sample was taken, followed by a third sample 40 min later. Immunocytochemistry One day after training, animals were sacrifi ced by transcardial perfusion with heparinized saline followed by fi xative, consisting of 4% paraformaldehyde in 0.1 M phosphate buffer. After dehydration of the brains in 30% buffered sucrose, 40 μm thick sections throughout the complete hippocampus, ranging from Bregma to -6.30, were cut on a cryostat microtome. In order to determine the effect of ASA acquisition on the survival of newly generated granule cells, animals were injected with the thymidine analogue BrdU one week before the start of the training. At this time point after BrdU-injection, labeled cells have stopped proliferating and are sensitive to undergo apoptosis (Dayer et al., 2003; Hastings and Gould, 1999), which enabled us to specifi cally investigate the effects of the learning task on the survival and not the proliferation of new cells (Prickaerts et al., 2004). Ki-67 immunocytochemistry was performed, to determine the effect of both ASA training paradigms on progenitor proliferation in the SGZ. Ki-67 is a nuclear protein, expressed in all mitotic cells throughout the cell cycle and is extensively used as a marker for proliferating cells (Kee et al., 2002). Sections that were reserved for BrdU-immunocytochemistry were collected in a cryoprotectant solution, containing 0.05 M phosphate buffer, 25% glycerol and 25% ethylene glycol, and stored at 20 C until immunocytochemistry was performed. The other series of sections were stored in 0.01 M PBS containing 0.1% sodium-azide. For both BrdU and Ki-67 immunocytochemistry, every twelfth section was taken. BrdU-immunocytochemistry required extra DNA denaturing steps (Kuhn et al., 1997). In brief, sections were exposed for 2 h at 65 C to 2x saline sodium citrate (2xSSC) containing 50% formamide, followed by a rinse with 2xSSC, 30 min incubation with 2 M HCl at 37 C and a washing step with 0.1 M borate buffer. Mouse-anti-BrdU and mouse-anti-ki-67 (both 1:200, Novocastra, Newcastle upon Tyne, UK) were used as the primary antibodies. Subsequently, sections for BrdU-immunocytochemistry were incubated in horse-anti-mouse (1:200, Vector Laboratories, Burlingame, USA), whereas goat-anti-mouse (1:400, Jackson, West Grove, USA) was used as the secondary antibody for the Ki-67 staining. Sections were incubated with HRP-conjugated streptavidin (Jackson) and the staining was visualized using diaminobenzidine (DAB, 20 mg/100 ml TBS) and 0.01% H 2 O 2. 68

70 Active shock avoidance learning and neurogenesis # BrdU-positive cells #BrdU-positivecells A B responders non-responders habituated controls home cage controls Figure 2: The number of BrdU-positive cells in the GCL of the DG. After one day (A) of ASA training no diff erences were observed between responders (n=9), non-responders (n=9) and habituated controls (n=8) in the number of BrdU-positive cells (one-way ANOVA: F(2,25)=0.70, P=0.51). Also four days of ASA training (B) did not induce any changes in the number of BrdU-positive cells (one-way ANOVA: F(3,29)=0.38, P=0.77). Responders: n=6, non-responders: n=4, habituated controls: n=10, home cage controls: n=10. C) Photograph of BrdU immunocytochemistry in the inner blade of the granule cell layer at 20x magnifi cation (upper panel). An enlargement of the selected region (40x magnifi cation) is shown in the lower panel. GCL= granule cell layer. C GCL Hilus 35 micron GCL Hilus 10 micron Quantifi cation of the immunocytochemistry Ki-67 and BrdU-positive cells in the SGZ, or one cell diameter deviating from this region, were counted throughout the focal plane using a 400x fi nal magnification. BrdU-positive cells in the GCL were included as well. The total number of counted cells was multiplied by twelve to obtain an estimation of the total number of positive cells per dentate gyrus (Cameron and McKay, 2001; Malberg et al., 2000). Statistics Behavioral data and CORT values were analyzed using a repeated measures ANOVA. Immunocytochemical data were analyzed using a one-way ANOVA. Whenever this revealed a significant outcome, pair wise comparison was performed with a Tukey-HSD test. Correlations between parameters were tested using a linear regression analysis. All data are expressed as means ± standard error of the mean (S.E.M.). 69

71 Chapter 5 Results Learning performance Animals were trained in a shuttle box for either one or four days and the number of active avoidances was recorded. Since in both experiments a considerable number of animals hardly displayed any active avoidance, it was decided to divide the animals in two groups: responders and non-responders. An animal was considered to be a responder if it was able to display four successive active avoidances. This criterion was reached by 50% of the animals that were trained for one day and by 60% of the rats that were trained during four days. Responders showed a clear learning curve with 78% (exp. 1, Fig. 1A) or 94% (exp. 2, Fig. 1B) active avoidances in the last block of 5 trials. Non-responders demonstrated only 18% (exp. 1, Fig. 1A) or 16% (exp. 2, Fig. 1B) active avoidances at the end of the training, which was significantly less than the responders (P<0.001 for both experiments) A #Ki-67positivecells Figure 3: The number of Ki-67 positive cells in the SGZ of the DG. After one day (A) of ASA training progenitor proliferation in the SGZ was not changed (one-way ANOVA: F(2,25)=0.38, P=0.69). Responders: n=9, non-responders: n=9, habituated controls: n=8. B) Four days of ASA training did not aff ect the number of BrdU-positive cells either (one-way ANOVA: F(3,29)=1.74, P=0.18). Responders: n=6, non-responders: n=4, habituated controls: n=10, home cage controls: n=10. C) Photograph of Ki-67 immunocytochemistry in the outer blade of the granule cell layer at 20x magnifi cation (upper panel). An enlargement of the selected region (40x magnifi cation) is shown in the lower panel. GCL=granule cell layer. # Ki-67 positive cells B C responders non-responders 35 micron GCL 10 micron habituated controls home cage controls Hilus GCL Hilus 70

72 Active shock avoidance learning and neurogenesis Eff ects of ASA on hippocampal neurogenesis The effects of ASA acquisition on the survival of newly produced granule cells were determined using BrdU immunocytochemistry. Neither the one-day (Fig. 2A) nor the four-day ASA protocol (Fig. 2B) had an effect on the survival of newly formed cells in the GCL compared to habituated and home cage controls. Moreover, no differences were observed between responders and non-responders. These results show that survival of newly generated hippocampal cells is not affected by either a short ASA protocol or a prolonged training. Furthermore, no correlation was observed between performance in the ASA task and the number of BrdU-positive cells (one-day ASA: R 2 =0.21, P=0.22, four-day ASA: R 2 =0.36, P=0.21). To determine the effect of both ASA training paradigms on progenitor proliferation in the SGZ, immunocytochemistry was performed for the mitotic marker Ki-67. Analysis of the Ki-67 staining revealed no differences between the experimental groups (Fig. 3A and 3B), indicating that neither one day nor four days of training in the ASA learning task influenced cell proliferation in the SGZ. CORT measurements In order to establish whether exposure to the shuttle box, either with or without CS and US presentation, increased baseline CORT levels, daily blood samples were taken from trained and habituated animals at the beginning of the day. On three of the experimental days, baseline samples were taken from home cage controls as well. No differences were observed between the groups. Moreover, in none of the groups changes were found in the course of the experiment (Fig. 4A), which is in line with previous studies on this topic (Coover et al., 1973; Knardahl and Murison, 1989). These data indicate that daily exposure to a shuttle box for four days does not induce long-term changes in basal HPA-axis activity. In addition to the baseline samples, blood was collected 20 min after the start of the training or habituation session to determine the height of the CORT peak. The average height of the CORT peak over the four training days did not differ between the groups (responders: 56 ± 2 μg/dl, non-responders: 48 ± 5 μg/dl, habituated animals: 49 ± 2 μg/dl, one-way ANOVA: F(2,19)=1.42, P=0.27). A third daily sample was taken 40 min later as a criterion for the recovery rate of the animals (Meerlo et al., 1999). From the baseline, peak and recovery samples an area under the curve was calculated as an estimate of the total CORT response. On none of the training days differences between the three groups were observed (Fig. 4B). Interestingly, the CORT response did not diminish over the four days, indicating that, in terms of HPA-axis reactivity, no habituation took place. Moreover, the CORT responses of the habituated control group that was allowed to freely explore the shuttle box without hearing the tone or receiving footshocks, were of a similar size as those of the rats that were exposed to CS and US. Importantly, since habituation to the shuttle box and ASA training evoked a clear but equal stress response, possible effects of ASA training on neurogenesis were not likely masked by stress. When the average CORT response of trained and habituated animals over the four experimental days was compared to the number of proliferating cells in the SGZ, a 71

73 Chapter 5 Plasma [CORT] (microg/dl) CORT response (AUC) #Ki-67positivecells day1 day2 day3 day4 day CORT response (AUC) responders non-responders habituated controls home cage controls day 1 day 2 day 3 day 4 A responders non-responders habituated controls B responders non-responders habituated controls C Figure 4: HPA-axis activity during the fourday ASA paradigm. Day 1 is the fi rst training day; day 5 is the day of perfusion. A) Baseline CORT levels did not diff er between the groups (repeated measures ANOVA, F(2,17)=0.54, P=0.59). B) On none of the training days differences could be observed in CORT response between the three groups (repeated measures ANOVA, F(2,17)=0.63, P= Data are presented as areas under the curve (μg*min/dl, AUC). C) After it had been determined that there were no diff erences in Ki-67 expression (one-way ANOVA: F(2,19)=1.63, P=0.23) or the average CORT response (one-way ANO- VA: F(2,19)=0.63, P=0.54), data of the three groups were pooled and regression analysis was performed. A signifi cant correlation was observed between the average CORT response over four days of ASA training or habituation to the shuttle box and the number of proliferating progenitors in the SGZ (R2=0.43, P=0.002). Responders: n=6, non-responders: n=4, habituated controls: n=10, home cage controls: n=10. significant correlation (R2=0.54, P=0.014) was observed, demonstrating that animals with high CORT responses had fewer Ki-67 positive cells in the SGZ than rats that displayed less HPA-axis reactivity upon ASA training or exploration of the shuttle box (Fig. 4C). 72

74 Active shock avoidance learning and neurogenesis Discussion In order to gain more insight into the connection between hippocampal neurogenesis and learning, particularly hippocampus-independent associative learning, rats were trained in a one-day or a four-day version of the ASA task. Progenitor proliferation and survival of newly generated granule cells were not changed after both ASA training paradigms. Blood samples taken throughout the four-day experiment revealed no changes in basal CORT levels. Although trained animals as well as habituated controls showed strong acute CORT responses upon exposure to the shuttle box that did not diminish in the course of the experiment, this repeated stress did not affect neurogenesis. On the other hand, a significant correlation was observed between the average CORT response over the four days and the number of proliferating progenitors in the SGZ. Several studies show that 40-60% of the newly formed granule cells die within one to three weeks after their formation (Dayer et al., 2003; Hastings and Gould, 1999). This substantial decrease of young granule cells was partially prevented when animals were exposed to MWM learning or TEC during this critical period (Gould et al., 1999a). The hippocampus-independent versions of the above-mentioned tasks were not able to induce changes in adult neurogenesis. ASA acquisition is not dependent on the hippocampus, but there is abundant evidence that the hippocampal formation is engaged in ASA learning. Training in the shuttle box lowered the threshold frequency to induce LTP in the DG (Ramirez and Carrer, 1989), increased the length of the postsynaptic density in the molecular layer (Van Reempts et al., 1992), and caused an increased immunoreactivity for muscarinic cholinergic receptors in the GCL starting 2 h after ASA training and lasting up to at least 24 h (Van der Zee and Luiten, 1999). In the current study, ASA training was started seven days after the fi rst BrdU injection, which overlaps with the time point at which the number of newly generated cells normally starts to decline. Our data show that neither one nor four days of ASA training influenced hippocampal neurogenesis. In line with previous fi ndings with hippocampus-independent learning tasks, these fi ndings suggest that only tasks that are dependent on the hippocampus may affect neurogenesis and that learning-induced activation of the hippocampal formation is not suffi cient to cause changes in the formation of new neurons. A comparable result was found with delay eyeblink conditioning, a hippocampus-independent task that, nevertheless, does activate certain regions of the hippocampal formation (Leuner et al., 2003). This task did not influence hippocampal neurogenesis either (Gould et al., 1999a; Shors et al., 2001). Furthermore, the fact that ASA acquisition induces changes in the hippocampal formation does not necessarily imply that the hippocampus has a stimulatory role in ASA learning. In fact, various data suggest that the presence of an intact hippocampus disturbs optimal ASA performance. Hippocampal lesions or damage of the fi mbria-fornix, for instance, facilitated ASA acquisition (Guenaire and Delacour, 1983; Guillazo-Blanch et al., 2002; Lovely, 1975; Weiner et al., 1998). Moreover, a significant correlation was observed between ASA learning and mossy fi ber distribution (Schwegler et al., 1981; Schwegler and 73

75 Chapter 5 Lipp, 1981). Mice and rats with low avoidance scores had more mossy fi bers terminating in the CA3 region. Granule cell density in the DG was shown to negatively correlate with ASA performance as well (Wimer et al., 1983). Perhaps, only those learning tasks in which the hippocampal formation positively contributes to the rate of acquisition result in an increased survival of newly formed granule cells. From a functional perspective it may be expected that only those cells will survive that actively participate in the acquisition process. Within the group of trained animals, a distinction was made between animals that had learned to avoid the foot shock and animals that had failed to show any avoidance response. The question can be raised how to consider the latter group, i.e. the non-responders. Especially during the initial phase of training, ASA can be regarded as a form of fear conditioning. This might explain the negative influence of the hippocampus in ASA performance, because fear-induced freezing behavior impedes acquisition of ASA (Gewirtz et al., 2000; Guillazo-Blanch et al., 2002). The non-responders may therefore have been exposed to fear conditioning only, which is interesting in view of the hippocampus-dependence of certain forms of fear conditioning. However, in the present study, it is not possible to distinguish between animals that only learned to associate the foot shock with the context, which is hippocampus dependent, and rats that did learn the relation between CS and foot shock, which is hippocampus-independent, but simply failed to show the avoidance response. The group of non-responders can therefore only be considered as a control group that received the same experimental treatment as the responders, but failed to show the correct behavioral response upon CS presentation. The number of cells expressing the mitotic marker Ki-67 was quantifi ed to determine the rate of hippocampal cell proliferation. In contrast to the survival of newly produced cells, there is no evidence that proliferation of hippocampal progenitors is influenced exclusively by hippocampus-dependent learning tasks. In fact, a recent study by Malberg and Duman (Malberg and Duman, 2003) showed a decrease in cell proliferation after conditioning using a shuttle box. In the present study, however, no differences were observed between the experimental groups. The difference between the studies can possibly be explained by the use of a different method that was used to determine the number of dividing cells. Ki-67 immunocytochemistry provides information about the number of mitotic cells at the moment of perfusion, which in the present study was 24 h after training. This means that acute or short-term changes in proliferation rate during the training process may have gone unnoticed, in contrast to the study of Malberg and Duman, where rats were injected with BrdU and perfused directly after training. It may thus be that ASA acquisition causes short-lasting changes in progenitor proliferation that could not be detected 24 hours after training. It is also possible that, similarly to fi ndings in the MWM (Dobrossy et al., 2003), cell proliferation changed dynamically throughout the four-day training protocol, without leading to a net alteration. Potential effects of ASA learning on hippocampal cell proliferation were not likely to be masked by effects of chronic stress induced by the experimental procedure. First, there were no persistent changes in baseline levels of CORT due to daily ASA training or 74

76 Active shock avoidance learning and neurogenesis habituation to the shuttle box. Second, there were no differences in acute stress responses between trained and habituated animals. Third, the acute stress response to training was similar in responders and non-responders. The fact that neurogenesis in undisturbed home cage control animals was similar to that in the experimental groups suggests that the shortterm elevations in plasma CORT during the four training days did not have a major impact on hippocampal neurogenesis. Yet, acute effects of ASA training on cell proliferation cannot be excluded, because Ki-67 staining identifi es cells that were proliferating at the moment of perfusion, a time point at which corticosterone levels had already returned to baseline. Nevertheless, a significant negative correlation was found between the CORT response of trained and habituated animals on the one hand and the number of proliferating progenitors in the SGZ on the other. Thus, even though the CORT responses did not directly affect Ki-67 expression, this correlation suggests that individual differences in HPA-axis reactivity are accompanied by differences in hippocampal cell proliferation. This hypothesis is supported by the fact that selection lines of rats (Lemaire et al., 1999) and mice (Veenema, 2003) that show differences in HPA-axis reactivity also differ with respect to hippocampal cell proliferation. The correlation between HPA-axis activity and hippocampal cell proliferation does not necessarily mean that there is a causal relationship between the two parameters. It can also indicate that HPA-axis reactivity and Ki-67 expression are variables that are influenced by the same intrinsic parameter. In summary, the present study shows that neither one day nor four days of ASA training affect hippocampal progenitor proliferation nor survival of newly formed cells in the DG granule cell layer. The current data confi rm that potential effects of learning on hippocampal neurogenesis might be dependent on multiple factors, such as the type of learning task the animals are trained in. Acknowledgements We thank Barbara Biemans and Theo Dinklo for their assistance with setting up the shuttle box and we thank Jan Bruggink for his contribution to the corticosterone assays. 75

77 76

78 6 Chapter Morris water maze learning in two rat strains increases PSA-NCAM expression in the dentate gyrus, but has no effect on hippocampal neurogenesis Karin Van der Borght 1, Alinde E. Wallinga 1, Paul G.M. Luiten 1, Bart J.L. Eggen 2, Eddy A. Van der Zee 1 1) Department of Molecular Neurobiology, Graduate school of Behavioural and Cognitive Neurosciences, and 2) Department of Developmental Genetics, Groningen Biomolecular Sciences and Biotechnology Institute; University of Groningen, P. O. Box 14, 9750 AA, Haren, The Netherlands Behavioral Neuroscience (2005), 119(4):

79 Chapter 6 Abstract The present study investigated whether Morris water maze learning induces alterations in hippocampal neurogenesis or NCAM polysialylation in the dentate gyrus. Two frequently used rat strains, Wistar and Sprague-Dawley, were trained in the spatial or the non-spatial version of the water maze. Both training paradigms had neither an eff ect on survival of newly formed cells that had been labeled seven to nine days prior to the training, nor on progenitor proliferation in the subgranular zone. However, the granule cell layer of the spatially trained animals contained signifi cantly more PSA-NCAM positive cells. These data demonstrate that Morris water maze learning causes plastic change in the dentate gyrus, without aff ecting hippocampal neurogenesis. 78

80 Morris water maze learning and neurogenesis Introduction The hippocampal dentate gyrus has, together with the olfactory bulb, the unique feature that it continues to produce new neurons during adult life (Altman, 1969; Altman and Das, 1965; Alvarez-Buylla and Garcia-Verdugo, 2002; Gross, 2000). The newly formed hippocampal neurons originate from undifferentiated progenitors that reside in the subgranular zone (SGZ) of the dentate gyrus (DG). Upon migration into the granule cell layer (GCL) they differentiate and become mature, functional granule cells (Cameron and McKay, 2001; Dayer et al., 2003; Hastings and Gould, 1999; Markakis and Gage, 1999; Van Praag et al., 2002). The regulation of adult hippocampal neurogenesis appears to be activity-dependent. Epileptic seizures in the dentate gyrus (Parent et al., 1997), amygdala kindling (Scott et al., 1998) or long-term potentiation in the mossy fi bers (Derrick et al., 2000) enhance proliferation of hippocampal progenitors in the subgranular zone. Increased behavioral activity, such as wheel running (Trejo et al., 2001; Van Praag et al., 1999b) and enriched housing (Kempermann et al., 1997; Nilsson et al., 1999), also stimulates hippocampal neurogenesis. Moreover, it has been reported that hippocampus-dependent learning tasks, such as the Morris water maze or trace eyeblink conditioning, have a positive effect on the formation of new neurons (Gould et al., 1999a). This effect seem to be specific for hippocampus-dependent learning tasks, since hippocampus-independent tasks, such as delay eyeblink conditioning or active shock avoidance learning, did not cause any changes in neurogenesis (Gould et al., 1999a; van der Borght et al., 2005a). It could be hypothesized that the activation of the hippocampal formation by certain types of learning can, at least partly, prevent the high level of cell death that normally occurs within two weeks after the generation of hippocampal granule neurons (Cameron and McKay, 2001; Dayer et al., 2003). However, using a somewhat different protocol, other researchers were not able to replicate these data for the Morris maze task in mice (Van Praag et al., 1999a), or they even found a decreased cell survival after spatial learning (Ambrogini et al., 2004b). Thus, spatial learning may affect hippocampal neurogenesis, but confl icting reports exist in the literature. Newly formed, immature hippocampal granule neurons express the polysialylated form of the neural cell adhesion molecule (PSA-NCAM) (Nakagawa et al., 2002; Seki and Arai, 1993). The presence of PSA-NCAM is generally associated with plastic changes in the central nervous system. It is abundantly expressed during development, where it mediates cell migration, neurite outgrowth and synaptogenesis (Edelman, 1986; Seki and Rutishauser, 1998). In adulthood, NCAM polysialylation is strongly reduced, but it appears to be upregulated in circumstances requiring structural remodeling (Ronn et al., 2000). Demyelination of the spinal cord, for instance, or hippocampal damage caused by epileptic seizures increase PSA-NCAM expression in the lesioned area (Dominguez et al., 2003; Oumesmar et al., 1995). PSA-NCAM has also been shown to be involved in learning as was shown by experiments in which PSA groups were removed from the NCAM molecule by 79

81 Chapter 6 treating rats with the enzyme endoneuraminidase NE (endo-n). This treatment resulted in impaired Morris water maze acquisition and retention (Becker et al., 1996). Moreover, different types of learning, like passive shock avoidance learning, Morris water maze training and contextual fear conditioning have been reported to stimulate NCAM polysialylation (Fox et al., 1995; Murphy et al., 1996; Sandi et al., 2003). The current study was aimed to investigate spatial learning-induced plastic changes in the dentate gyrus in relation to neurogenesis. Since the potential effect of hippocampusdependent learning on adult neurogenesis is still debated, we investigated whether Morris water maze learning in rats affects survival of newly formed cells and proliferation of hippocampal progenitors in the dentate gyrus. We also analyzed PSA-NCAM expression in the dentate gyrus to relate learning-induced changes in NCAM polysialylation to potential alterations in hippocampal neurogenesis. Because it is known that learning performance and hippocampal neurogenesis differ significantly between inbred laboratory mouse strains (Kempermann and Gage, 2002a; Kempermann and Gage, 2002b), we decided to compare learning capacity, baseline neurogenesis and learning-induced changes in plasticity in the dentate gyrus between two widely used rat strains, Wistar and Sprague-Dawley. Materials and Methods Animals and housing Twenty-four male Wistar rats (338 ± 24 g, bred in our own facilities) and twenty-one male Sprague-Dawley rats (336 ± 30 g, Harlan, Horst, The Netherlands) were individually housed. The animals had free access to water and food and were kept under a 12:12h light: dark cycle, lights on at 7:00 a.m. All procedures concerning animal care and treatment were in accordance with the regulations of the ethical committee for the use of experimental animals of the University of Groningen (DEC number 2719). Morris water maze training and BrdU injections The Morris water maze consisted of a black pool (diameter 140 cm) fi lled with water (26 ± 1 C). A small, black platform (diameter 9 cm) was placed at 23 cm from the border of the pool and 2.5 cm under the water surface in order to make it invisible to the animals. The behavior of the animals in the pool could be tracked with a camera connected to a computer. Specialized software (Ethovision, Noldus, Wageningen, The Netherlands) allowed us to measure various parameters, such as swim speed, the distance moved and the latency to fi nd the platform. Place learners (Wistar: n=8, Sprague-Dawley: n=7) were trained using a protocol of fi ve trials per day, with an intertrial interval of twenty minutes, for fi ve consecutive days. Animals were allowed to swim for maximally 60 seconds per trial. The fi rst trial of the fi rst day was performed without a platform, in order to give the animals the opportunity to habituate to the swimming procedure. In the second trial of the fi rst day, the platform was 80

82 Morris water maze learning and neurogenesis present in the maze. If the animals had not been able to fi nd the platform within 60 seconds, they were guided there by the experimenter. After having reached the platform, animals were kept there for 10 seconds to give them the opportunity to orientate themselves to the spatial cues that were present in the experimental room. The platform was kept in the same position for three days, though the starting position of the animals was changed between trials. After three days, the platform position was changed and the animals had to learn the new position of the platform on training days 4 and 5. Two control groups were included in the experiment: home cage controls (Wistar: n=8, Sprague-Dawley: n=7) and cue learners (Wistar: n=8, Sprague-Dawley: n=7). The cue learners underwent the same procedure as the place learners, except for the fact that the spatial learning component was lacking. The platform was made visible to the animals by placing it 1 cm above the water surface, by making it white-colored and by placing a fl ag on it. In every trial, the platform was placed in a different position. Home cage controls remained undisturbed throughout the experiment. Seven to nine days before the start of the training, all animals were intraperitoneally injected with 100 mg/kg BrdU (Sigma, St. Louis, USA) dissolved in saline (20 mg/ml) once a day for three consecutive days. Brain processing and immunocytochemistry One day after training, approximately hours after the last training session, animals were sacrificed by transcardial perfusion with heparinized saline, followed 2.5% paraformaldeyde and 0.05% glutardialdehyde in 0.1 M phosphate buffer (PB). After dehydration in 30% sucrose, 40 μm coronal sections were cut on a cryostat microtome. Twelve series spanning the entire hippocampus (Bregma to Bregma -6.30) were collected in cryoprotectant (0.05 M PB, 25% glycerol and 25% ethylene glycol) and stored at -20 C until they were used for immunocytochemistry. BrdU and Ki-67 immunocytochemistry were performed on every twelfth section of the hippocampus, using a protocol as described earlier (van der Borght et al., 2005a). In brief, sections for the BrdU staining underwent some extra steps for DNA denaturation. For this purpose, they were exposed to 50% formamide in 2XSSC at 65 C and 0.2 M HCl at 37 C. The primary antibodies that were applied were rat-anti-brdu (1:800, Oxford Biotechnology, Oxfordshire, UK) and mouse-anti-ki-67 (1:200, Novocastra, Newcastle upon Tyne, UK). As secondary antibodies, biotinylated donkey-anti-rat and biotinylated sheepanti-mouse (both 1:200, Jackson, West Grove, USA) were used. Staining was visualized with diaminobenzidine (20 mg/100 ml, DAB) as chromogen. For the PSA-NCAM staining, fi ve to six representative sections from the dorsal hippocampus were selected. After preincubation with 3% normal rabbit serum and 0.5% triton-x100 they were incubated with the primary antibody (1:1000, mouse-anti PSA- NCAM IgM, Chemicon, Temecula, USA) for 96h. As a secondary antibody, rabbit-antimouse IgM (1:200, Jackson) was used. After incubation with the ABC kit (Vector, Burlingame, UK), staining was visualized with DAB. 81

83 Dstancet i o platf o rm ( cm) Distance topl atfom(cm) r Chapter 6 Figure 1: Learning curves of Wistar (n=8) and Sprague-Dawley (n=7) rats in the place (A) or cue (B) version of the Morris water maze. Training consisted of fi ve trials per day for fi ve consecutive days. The fi rst trial on day 1 was performed without a platform. In the group of place learners, the platform was relocated after three days of training. Both rat strains performed equally well in the place learning task, but Wistar performed signifi cantly better than Sprague-Dawleys during cue learning (P=0.001). Data are expressed as mean distance moved before reaching the platform ± S.E.M Wistar Sprague-Dawley Wistar S prague-dawley Trial Trial A B Quantifi cation During the analysis of the brain material, the experimenter was blind to the treatment of the animals. BrdU and Ki-67 immunopositive cells were counted in every twelfth section of the hippocampal formation with a 40x objective. Only cells that were in the subgranular zone or one cell diameter deviating from this region were included. BrdU-positive cells that were lying in the granule cell layer were counted as well. The number of counted cells was multiplied by twelve to get an estimation of the total number of positive cells per dentate gyrus. For the PSA-NCAM staining, all cells in the subgranular and granular layer were counted in fi ve to six sections that were randomly chosen to be representative for the dorsal hippocampus. The average cell number per section was calculated. Statistics Morris Water maze behavioral data were analyzed with a repeated measures ANOVA. When three experimental groups were compared, BrdU, Ki-67 and PSA-NCAM cell counts were statistically tested with a one-way ANOVA. If this revealed a significant outcome, a Bonferroni test was applied for post hoc testing. Comparison between two groups was performed using an independent-samples t-test. 82

84 # BrdU-positive cel ls Morris water maze learning and neurogenesis A B Wistar * Sprague-Dawley Home cage Cue learners Place learners Figure 2: The number of BrdU-positive cells per dentate gyrus. A) Neither in Wistars (n=8 per group) nor in Sprague-Dawleys (n=7 per group) any diff erence was observed in BrdU-positive cell number between home cage controls, cue learners or place learners. However, Wistar home cage controls had signifi cantly more BrdU-positive cells than the Sprague-Dawley home cage control animals (* P<0.01). Data are expressed as mean ± S.E.M. B) Representative photomicrographs of BrdU-immunocytochemistry. Scale bar = 50 μm in the upper panel. A magnifi cation of the selected region is shown in the lower panel (scale bar = 10 μm). Results Behavioral testing The two rat strains performed equally well in the spatial version of the Morris water maze, in which they had to fi nd the hidden platform (Fig. 1A, Between strains: F(1,13)=0.10, P=0.75, strain x trial: F(13,169)=0.70, P=0.76). After relocation of the platform on day 4, the animals quickly learned to fi nd the new position of the platform. Also in this reversal learning paradigm, no differences were observed between Wistar and Sprague-Dawley (Fig. 1A, Between strains: F(1,13)=1.53, P=0.24, strain x trial: F(9,117)=0.86, P=0.57). As expected, the rats showed a decrease in the distance they needed to swim to fi nd the platform (P<0.001 for both the fi rst fi fteen trials and the last ten trials). The latency to fi nd the platform could not be used as an indicator of learning performance, since the two strains significantly differed in swim speed (Wistar: 18.5±0.6 cm/s, Sprague-Dawley: 23.2±0.4 cm/s, F1,29=32.99, P<0.001). Therefore, the distance swum by the rats until they reached the platform was taken. In the cued version of the Morris water maze, the rats acquired the task rapidly (Fig. 1B, P<0.001). Sprague-Dawley rats swam a greater distance before reaching the platform than Wistars (Fig. 1B, Between strains: F(1,13)=20.27, P=0.001), but both strains managed to acquire the task. Moreover, there was no significant interaction between strain and trial (F(23,299)=0.93, P=0.56). Learning speed differed significantly between place learners and rats that were trained with the visible platform (Wistar: P<0.001, Sprague-Dawley: P<0.01). 83

85 Chapter 6 #Ki-67positivecells A B Wistar Sprague-Dawley Home cage Cue learners Place learners Figure 3: Ki-67 expression in the hippocampal subgranular zone. A) Neither hippocampus-independent nor hippocampus-dependent learning in the water maze caused a change in hippocampal cell proliferation. This was the case for both rat strains (Wistar n=8 for all groups, Sprague-Dawley: n=7 for all groups). Also, no strain diff erences were observed in Ki-67 expression. Data are expressed as mean ± S.E.M. B) Example of Ki-67 immunocytochemistry in the hippocampus (scale bar = 50 μm). The insert shows an enlargement of the selected region (scale bar = 10 μm). BrdU In order to investigate the effects of the learning task on the survival of newly formed hippocampal cells, rats were injected with the thymidine analogue BrdU seven to nine days before the start of the training. One day after the last training, animals were sacrifi ced and brains were processed for immunocytochemistry. Quantification of the number of BrdU-positive cells in the dentate gyrus did not reveal any differences between home cage controls, cue learners and place learners (Fig. 2). This was the case for both rat strains (Wistar: F(2,23)=1.49, P=0.25; Sprague-Dawley: F(2,20)=0.032, P=0.97). These data indicate that Morris water maze learning did not promote survival of newly generated cells in the hippocampus. However, a significant difference was observed in the number of BrdU-positive cells between home cage controls of the two rats strains, with Sprague-Dawleys having 42% less positive cells than Wistars (F(1,14)=28.94, P<0.01). Ki-67 The Ki-67 protein is expressed in all cells during all phases of the cell cycle, except G0 (Scholzen and Gerdes, 2000) and can therefore be considered as a good indicator for the number of proliferating cells that were present at the moment of perfusion. Quantifi cation of the number of Ki-67 positive cells in the subgranular zone showed that neither place learning nor cue learning caused a change in hippocampal cell proliferation (Fig. 3, Wistar: F(2,23)=0.78, P=0.47; Sprague-Dawley: F(2,20)=0.99, P=0.39). Also, Ki-67 expression did not differ between the home cage controls of both strains (F(1,14)=1.40, P=0.26), indicating that baseline hippocampal cell proliferation is similar for Wistars and Sprague-Dawleys. 84

86 # PSA-NCAM pos itive cells/section Morris water maze learning and neurogenesis PSA-NCAM The binding of α2,8-linked polysialic acid homopolymers to the neural cell adhesion molecule (PSA-NCAM) has been associated with plastic changes in the brain. Moreover, PSA- NCAM is expressed by immature neurons in the adult hippocampus. Analysis of PSA- NCAM immunocytochemistry showed a signifi cant effect of place learning. In the Wistar rats, place learners had 19% more PSA-NCAM positive cells than home cage controls (Fig. 4, P<0.05). Also in the Sprague-Dawleys a learning effect was observed. Place learners had 31% more immunoreactive cells compared to home cage controls (P<0.001). Moreover, comparison between the two strains, with regard to baseline PSA-NCAM expression in home cage controls, showed that Wistar rats had 40% more PSA-NCAM positive cells in the dentate gyrus than Sprague-Dawleys (P<0.001) A B ** * ** ** Wistar Sprague-Dawley Home cage Cue learners Place learners Figure 4: PSA-NCAM expression in the hippocampal subgranular zone. A) Both in Wistars and Sprague-Dawleys a learning-induced increase in PSA-NCAM immunoreactivity in the subgranular zone was observed. In Sprague-Dawleys, place learners also had signifi cantly more PSA-NCAM positive cells than cue learners. Moreover, when comparing home cage control animals, a signifi cant diff erence was observed between Wistars and Sprague-Dawleys, with Wistars having more PSA-NCAM immunopositive cells than Sprague-Dawleys. * P<0.05, ** P< Data are expressed as mean ± S.E.M. B) Photomicrograph of PSA-NCAM immunocytochemistry in the dentate gyrus. Scale bar = 50 μm. 85

87 Chapter 6 Discussion The present study investigated the occurrence of plastic changes in relation to neurogenesis in the hippocampal dentate gyrus following training in a spatial learning task, the Morris water maze. The data show that place learning in the water maze induced an increased expression of PSA-NCAM. Hippocampal progenitor proliferation and survival of newly formed cells were not altered by the spatial learning task. The literature on spatial learning-induced changes in newly formed hippocampal cell survival is not entirely consistent. Between one and three weeks after their formation, a large part of the newly formed granule cells die (Cameron and McKay, 2001; Dayer et al., 2003; Hastings and Gould, 1999). Gould and colleagues (1999) reported that training rats in a spatial learning task within this critical period, that is, starting seven days after injection with BrdU, could prevent many newly formed cells from undergoing apoptosis. In contrast, others observed a negative effect on survival of newly generated hippocampal cells, when starting Morris water maze training eight to ten days after BrdU administration (Ambrogini et al., 2004b). In the present study, in which water maze training was started seven to nine days after BrdU injections, no effects on survival of BrdU-labeled cells could be demonstrated. Possibly, the time window in which the effects of learning on newly formed cell survival are investigated is very narrow. At the age of 10 days, only 9% of the cells has formed axons towards the CA3 region (Hastings and Gould, 1999), which reduces the possibility that the BrdU-labeled cells in the present study actively participated in the learning process and that this participation could rescue them from going into apoptosis. Moreover, other experimental approaches in which neurogenesis was partially ablated by treatment with antimitotic drugs (Shors et al., 2001; Shors et al., 2002)) or by cranial irradiation (Madsen et al., 2003; Snyder et al., 2005) did not result in an impairment in Morris water maze learning. These studies suggest that hippocampal neurogenesis is not required for Morris water maze learning, which minimizes the likelihood that water maze learning stimulates hippocampal neurogenesis. Our data also indicated that Morris water maze learning had no effect on cell proliferation in the hippocampal subgranular zone. This fits with other reports (Gould et al., 1999a; Van Praag et al., 1999b), although there is also evidence for an increase in hippocampal cell proliferation after Morris water maze learning (Lemaire et al., 2000). A recent study by Dobrossy and colleagues (Dobrossy et al., 2003) demonstrated that Morris water maze learning does not result in a net change in cell proliferation, but that cell proliferation is increased during the initial phase of the learning process, and that these newly formed cells die during the late phase of the learning process. In the current study, cell proliferation was determined on the basis of Ki-67 expression, which only provides information about the number of proliferating cells at the moment of perfusion, which was one day after the last training. Dynamic changes during the learning process can therefore not be excluded. Yet, our data suggest that hippocampus-dependent learning does not cause long-term changes in hippocampal cell proliferation. 86

88 Morris water maze learning and neurogenesis The increase in the number of cells that express the polysialylated form of NCAM eighteen hours after training is in line with earlier studies (Fox et al., 1995; Murphy et al., 1996; Sandi et al., 2003) and it indicates that the learning task induced plastic changes in the dentate gyrus. PSA-NCAM is mainly observed in the subgranular zone of the dentate gyrus, the site of hippocampal neurogenesis and it is also expressed by newly formed cells that are one to three weeks old (Nakagawa et al., 2002; Seki, 2002a; Seki, 2002b; Seki and Arai, 1993; Seki and Arai, 1999). However, since our data show did not show any changes in BrdU-positive cell number after learning, the learning-induced increase in NCAM polysialylation is probably associated with plastic changes in the dentate gyrus, such as neurite outgrowth, dendritic branching or modifi cation of intracellular signaling cascades (Muller et al., 1996; Rutishauser et al., 1988), but not with alterations in hippocampal neurogenesis. Finally, under baseline conditions, i.e. in home cage control animals, Wistar rats had significantly more BrdU-positive cells in the granule cell layer than Sprague-Dawleys. Because the Ki-67 staining showed that the production of new cells was similar for the two strains, it can be suggested that, within the time window that was investigated, a higher percentage of newly formed cells had died in Sprague-Dawley rats. The strain difference in BrdU-positive cell number was reflected in PSA-NCAM expression, which is expressed by immature neurons. The strain-dependent difference in hippocampal neurogenesis had no impact on performance in the Morris water maze, which reduces the probability of a direct relation between the formation of new granule neurons and hippocampus-dependent learning. In summary, we demonstrate a spatial learning-induced increase in NCAM polysialylation in the dentate gyrus, but no effect on hippocampal neurogenesis. These data show that behavioral interventions that induce plastic changes in the hippocampal formation are not suffi cient for inducing alterations in hippocampal neurogenesis. 87

89 88

90 7 Chapter Memory retrieval reduces the number of newly formed hippocampal neurons in mice with baseline and exercise-enhanced levels of neurogenesis Karin Van der Borght1, Robbert Havekes1, Thomas Bos1, Bart J.L. Eggen2, Eddy A. Van der Zee1 1) Department of Molecular Neurobiology, Graduate school of Behavioural and Cognitive Neurosciences, and 2) Department of Developmental Genetics, Groningen Biomolecular Sciences and Biotechnology Institute; University of Groningen, P. O. Box 14, 9750 AA, Haren, The Netherlands Submitted 89

91 Chapter 7 Abstract Adult hippocampal neurogenesis has been hypothesized to play a role in the formation and consolidation of memory. It has also been suggested that the integration of new neurons into the hippocampal network removes outdated information from the hippocampus. Here, we tested both theories using a 14-day exercise paradigm, which caused a signifi cant increase in hippocampal neurogenesis. First, we investigated whether enhanced hippocampal neurogenesis improved performance in the Y-maze, a spatial hippocampus-dependent learning task. Our data show that animals with elevated levels of hippocampal neurogenesis acquired this task signifi cantly faster than sedentary controls. Next, we tested the hypothesis that enhanced neurogenesis removes old memories from the hippocampus. Mice were trained in the Y-maze and hippocampal neurogenesis was subsequently stimulated by 14 days of wheel running. After the exercise period, we observed an improved performance in memory retention and reversal learning, indicating that retrieval of spatial information was facilitated, instead of impaired by enhanced neurogenesis. Therefore, these fi ndings do not support the memory clearance hypothesis. Moreover, memory retrieval itself caused changes in the generation of new neurons. We observed a signifi cant decrease in the number of newly formed doublecortin-immunoreactive neurons and pcreb-positive cells in the subgranular zone after memory retention and reversal learning. This decrease was observed in both sedentary controls and animals with enhanced neurogenesis. We hypothesize that the formation of new hippocampal neurons is transiently decreased during memory retrieval, in order to reduce interference between memory formation and the retrieval of existing memories. 90

92 Memory retrieval reduces neurogenesis Introduction There is increasing evidence that newly formed granule neurons in the dentate gyrus play a role in learning and memory processes. Hippocampus-dependent learning can promote the survival of newly formed granule neurons (Ambrogini et al., 2000; Gould et al., 1999a), which suggests that the young neurons have a function in memory formation. Furthermore, inhibition of neurogenesis, by cytostatic drugs or brain irradiation, showed that newly formed neurons are crucial for certain types of hippocampus-dependent learning (Shors et al., 2001; Shors et al., 2002) and for memory consolidation (Bruel-Jungerman et al., 2005; Snyder et al., 2005). There is also evidence that the incorporation of new neurons in the hippocampal circuitry results in the clearance of outdated information. Computer models of a simple network predict that turnover of cells accelerates removal of information (Chambers et al., 2004; Deisseroth et al., 2004). There is in vivo data evidence supporting this theory. Presenilin-1 (PS1) knockout mice do not show an increase in hippocampal neurogenesis upon cage enrichment. Enriched housing of PS1 knockout mice following acquisition of a contextual fear-conditioning task, resulted in a better memory retention than wildtype controls (Feng et al., 2001). These data suggest that the elevated levels of hippocampal neurogenesis in the wildtype animals promoted removal of the hippocampal memory trace. In the present study, we further investigated the function of newly generated neurons in memory formation and memory retrieval. We hypothesized that if reduced levels of neurogenesis cause learning and memory impairments, increased numbers of newly formed hippocampal granule neurons should facilitate the acquisition of a learning task. Additionally, we predicted that if neurogenesis has a function in the deletion of old information from the hippocampus, this clearance process can be accelerated by an increased production of new neurons. Hippocampal neurogenesis was stimulated by housing mice with a running wheel for 14 days (Aberg et al., 2003; Kronenberg et al., 2003; Van Praag et al., 1999b), which is sufficient to give the newly formed neurons the opportunity to mature and to functionally integrate (Hastings and Gould, 1999; Kempermann et al., 2004). First, we determined the effects of enhanced neurogenesis on the acquisition of the Y-maze, a hippocampus-dependent (Bannerman et al., 2003) spatial learning task. Second, mice were trained in the Y- maze and subsequently housed with a running wheel for 14 days. After this period memory retention and reversal learning performance were tested. Here, we show that enhancement of neurogenesis by wheel running improved learning in the Y-maze. In contrast to what would be expected on the basis of the memory clearance hypothesis, memory retrieval was also facilitated by exercise. These data suggest that newly generated neurons do not remove information from the hippocampus, but seem to play an important role in retrieval. Therefore, we further studied the effects of memory 91

93 # D CX -p ositiv e cells Chapter A B ** exercise P standard cage P Food restriction Figure 1: Eff ect of exercise on hippocampal neurogenesis and CREB phosphorylation. A) Overview of the experimental protocol; p = perfusion; numbers indicate days. B) The 14-day exercise procedure resulted in a signifi cant increase in the number of DCX-positive cells. Cell numbers are expressed as the average number of DCX-positive cells per section. C) Runners show signifi cantly more pcreb-positive cells in the granule cell layer than sedentary animals. **: P<0.01. Cell numbers are shown as average number of cells/ section ± S.E.M. 0 Sedentary Exercise % DCX -p osi tve i den dr it es C *** sedentary exercise # pcre B -positive cells D ** 0 Sedentary Exercise retention and reversal learning on hippocampal neurogenesis by staining the brains for doublecortin and pcreb, both of which are markers that are expressed by newly formed neurons. Here, we show that both memory retention and reversal learning reduced hippocampal neurogenesis. 92

94 Percent agecorrect arm e nt ries Memory retrieval reduces neurogenesis Figure 2: Eff ect of exercise on Y-maze acquisition. A) Schedule of the experimental protocol that was used. The upper line shows the experimental procedure for the enhanced neurogenesis group, the lower one represents the sedentary mice; sc = standard cage, hab = habituation to the Y-maze, p = perfusion; numbers indicate days. B) The learning curves show that animals with enhanced neurogenesis performed signifi - cantly better in the Y-maze than sedentary mice. The learning curves show the average percentage of correct arm choices per session of six trials ± S.E.M. C) Since the exercise animals performed above chance level during the fi rst session, this session was subdivided into the individual trials. The graph shows that runners did not have an initial preference for the baited arm, but that they quickly learned which arm contained the food. The learning curves show the average percentage of mice that entered the correct arm per trial ± S.E.M. A Sedentary controls Enhanced neurogenesis exercise B C standard cage sc hab hab Food restriction Session training training P 21 P Trial Materials and Methods Animals and housing conditions Eighty male C57Bl/6 mice (Harlan, Horst, The Netherlands, 25.8 ± 0.2 g at the beginning of the experiment) were individually housed, had free access to drinking water and were kept under a 12/12h light/dark cycle (lights on at 8.00 a.m.). Throughout the entire experiment, all animals were food restricted to about 85% of their original bodyweight, which means that they received on average 3 to 4 g of food per day. Animals were fed daily between 3:00 p.m. and 5:00 p.m., after Y-maze training. Food restriction started 1 day prior to the start of the experiment. All procedures concerning animal care and treatment were approved by the ethical committee for the use of experimental animals of the University of Groningen (DEC 4089A and 4089B). 93

95 Perc enta ge corr ect arm ent re i s Percen tage corr ect arm entri es Chapter 7 Y-maze training procedure Behavioral testing was conducted in an enclosed plexiglass Y-maze. The home cage of each animal was provided with a small sliding door that could be connected to the maze. Both the stem arm (27.5 cm long) and the two arms forming the Y (both 27.5 cm long and diverging at a 60 angle from the stem arm) were 5 cm in diameter. Perforations at the endings of the two arms forming the Y allowed odors from food (standard lab chow, Hope Farms, Woerden, The Netherlands) placed under the perforations to enter both arms. Small plastic blocks (1 cm high) were placed 4 cm from the endings of the arms to prevent visual inspection for food presence from a distance. Each arm was equipped with a trapdoor halfway the arm which could be operated manually from the experimenter s position. The day prior to the start of the training, animals were allowed to freely explore the maze for fi ve minutes. Next, they received two trials, one in which the food was located in the left arm and one in which the food was positioned in the right arm. This procedure prevented the development of a preference for one of the arms. During the training procedure, the animal voluntarily entered the maze and whenever it visited one of the two A hab training exercise sc retention/reversal P arms, the trapdoor of the non-visited arm was hab training standard cage retention/reversal P Figure 3: Eff ect of exercise on memory retention and reversal learning. A) Schematic overview of the experimental procedure. The upper line represents the experimental protocol for animals with enhanced neurogenesis, the lower line show the procedure for sedentary mice; hab = habituation to the Y-maze, sc = standard cage, p = perfusion; numbers indicate days. B) All animals were fully trained in the Y-maze (grey squares). Directly after the last training session, mice were either housed with a running wheel or kept in a standard cage for 14 days. After this 14-day interval, which is shown by a break in the x- axis, animals with enhanced neurogenesis performed signifi cantly better in the retention test compared to sedentary controls. C) Also in the reversal task, in which the food was located in the arm opposite to the one during training, mice with enhanced neurogenesis showed an improved performance compared to sedentary mice. The learning curves show the average percentage of correct arm choices per session of six trials ± S.E.M All mice together Sedentary Enhanced neurogenesis Session Reversal learning B C Food restriction Allmicetogether Sedentary Enhanced neurogenesis Session 24 Memory retention 94

96 Memory retrieval reduces neurogenesis closed. The mouse was allowed to eat the food crumble and after it had re-entered its home cage, the arm connected to the home cage was closed. After thorough cleaning of the arms, the animal was allowed to enter the maze again for the next trial. Eff ect of 14 days of exercise on hippocampal neurogenesis Eight mice were housed with a running wheel (diameter 13 cm) for 14 days. Sedentary mice (n=6) were housed under standard conditions. Running wheel activity was recorded and analyzed with specialized software (ERS system, Haren, The Netherlands). After the exercise period, mice were sacrifi ced and brains were processed for immunocytochemistry (see below). Eff ect of enhanced neurogenesis on memory formation Mice (n=16) were housed with a running wheel for 14 days. Sedentary mice (n=16) were kept in a standard cage during this period. After the exercise period, all mice were placed in a clean, standard cage and one day later eight exercise animals and eight sedentary animals were habituated to the Y-maze. Training took place on the following four days and consisted of two sessions per day, each session containing six trials. The other half of the exercise and the sedentary animals were kept in the home cage during the training period and served as naive controls for the Y-maze. Animals were sacrifi ced one day after the last training session. Eff ect of enhanced neurogenesis on memory retention and reversal learning Mice (n=34) were trained in the Y-maze for three days, using two sessions per day. Directly after the last training session, 16 animals were housed with a running wheel for 14 days and the rest of the mice remained in a standard cage during this period. One day after the end of the exercise period, all mice were placed in a clean, standard cage for one day. The next day, memory retention was tested in half of the exercise animals and half of sedentary mice. These animals were placed again in the Y-maze, with the food in the same arm as they had learned during training. The other half of the animals was tested in a reversal learning task, in which the food was located in the arm that was not baited during the training sessions. Animals were exposed to the retention or reversal learning paradigm for two sessions per day for four days and were sacrificed one day later. Brain processing and Immunocytochemistry Animals were transcardially perfused with heparinized saline, followed by 4% paraformaldehyde in 0.1 M phosphate buffer. Brains were removed, kept in 0.01 M PBS overnight and subsequently cryoprotected in 30% sucrose for 48h. Next, 30 μm sections, spanning the dorsal hippocampus (Bregma to -2.80), were cut on a cryostat microtome. Brains were stained for doublecortin (DCX) and Ser133-phosphorylated CREB (pcreb). Sections were treated with 0.3% H 2 O 2, blocked with 3% normal serum and the cell membrane was permeabilized with 0.1% Triton-X100. Goat-anti-DCX (1:1000, Santa Cruz Biotechnology, Santa Cruz, CA, USA) and rabbit-anti-pcreb (1:300, Cell Signaling Technology, 95

97 Chapter 7 Beverly, MA, USA) were applied for 72h at 4 C. Sections were incubated with a biotinylated secondary antibody (rabbit-anti-goat or goat-anti-rabbit, 1:400, Jackson Immunolabs, West Grove, PA, USA) for 2h at room temperature, followed by incubation with Avidin- Biotin-Complex (1:400, ABC Elite kit, Vector Laboratories, Burlingame, CA, USA) for 2h. Staining was visualized with 20 mg/100 ml DAB and 0.03% H 2 O 2. For the DCX/pCREB double labeling procedure, a pcreb staining was performed as described above, but without the use of normal serum. The staining was visualized with DAB (20 mg/100 ml), nickelammoniumsulfate (200 mg/100 ml) and 0.03% H 2 O 2. Subsequently, sections were treated with a high dose of H 2 O 2 (1%), in order to remove all HRP reactivity. Then, sections were stained for DCX, as described above. The DCX-positive cells were visualized with DAB (15 mg/100 ml) and 0.01% H 2 O 2. Quantifi cation of the immunostainings DCX and pcreb-positive cell numbers were determined in 3-4 sections per animal, randomly chosen between to mm from Bregma. Cells were counted throughout the entire thickness of the sections, using a 40x magnifi cation. In order to prevent the inclusion of cell profi les in the DCX analysis, only cells were included with a cell soma that was larger than 8 μm in diameter. The inner and outer blades of the granule cell layer were counted separately. For every animal, the average number of immunopositive cells/section was calculated. A second measure of the DCX-immunostaining was taken in order to verify the DCX cell count analysis. Since most DCX-immunopositive cells possess only one primary dendrite that projects through the granule cell layer, we determined the density of DCXimmunoreactive dendrites in the granule cell layer as a measure for the number of DCXpositive cells. For the density measurements the same sections were used as for the cell counts. With a computerized system (Leica Qwin, Rijswijk, The Netherlands), two equally sized areas of the inner blade of the granular cell layer and two areas of the outer blade of each hippocampus were delineated. Within the demarcated areas, the percentage of the total surface that was covered with immunopositive dendrites was calculated. For every animal, an average area percentage was calculated from the different measurements. Statistics The learning curves of the Y-maze were analyzed using a repeated measures ANOVA. Potential differences in DCX and pcreb immunoreactivity after 14 days of exercise were statistically tested with an independent-samples t-test. For the statistical analysis of the DCX-measurements and the pcreb data of the different Y-maze groups, a two-way ANO- VA was used, with exercise and Y-maze training as between-subjects variables. Whenever this revealed a significant difference, pairwise comparisons were made with a post hoc LSD test. Data are expressed as averages ± S.E.M. 96

98 Memory retrieval reduces neurogenesis Results Running wheel activity in combination with dietary restriction promotes hippocampal neurogenesis This study aimed to look at the effects of enhanced neurogenesis on various aspects of Y- maze learning. Running wheel activity has repeatedly been shown to promote hippocampal neurogenesis. In the experiments reported here, exercise was combined with a food restriction paradigm. We examined whether neurogenesis is still enhanced by exercise under conditions of reduced food intake. Animals were housed with a running wheel for 14 days and perfused one day later. An overview of the experimental procedure that was used is presented in Fig. 1A. DCXimmunocytochemistry showed that the number of immature neurons was signifi cantly increased by the exercise procedure (Fig. 1B, P 0.01). This increase was visible both in the inner and the outer blade of the granule cell layer (separate data sets not shown). The difference between runners and control mice was confi rmed by the DCX-positive dendrite measurements (Fig. 1C, P 0.001). In addition, brains were stained for pcreb. Figure 1D shows that 14 days of exercise significantly increased the number of pcreb-positive cells in the DG (P 0.01), and this was observed in both blades of the GCL (separate data sets not shown). Running wheel activity facilitates Y-maze acquisition, retention and reversal learning We investigated the consequences of enhanced neurogenesis caused by 14 days of voluntary exercise on various aspects of Y-maze learning. First, the effects of wheel running on acquisition of the Y-maze were studied (Fig. 2A). Animals were exposed to two training sessions per day and each session consisted of six trials. Overall, the runners performed significantly better than sedentary mice (Fig. 2B, repeated measures ANOVA: F(1,14)=30.1, P 0.001; session*group interaction: F(7,98)=3.88, P 0.001). A more detailed examination of the fi rst session (Fig. 2C) shows that the mice with enhanced neurogenesis did not have an initial bias for the correct arm, but readily learned the position of the food within the fi rst training session (repeated measures ANOVA: F(1,14)=13.5, P 0.01). Next, we investigated whether 14 days of exercise, starting directly after Y-maze training was completed, had an effect on memory retention. This was tested by re-exposing the animals to the maze after the 14-day exercise period and to test their memory for the position of the food (Fig. 3A). Animals quickly relearned the position of the food (repeated measures ANOVA, session effect: P 0.05). Furthermore, there was a signifi cant session*group interaction (F(7,105)=2.4, P 0.05), indicating that running wheel activity facilitated Y-maze retention (Fig. 3B). Third, the effects of running wheel activity on reversal learning were tested. A similar protocol was used as for the retention test (Fig. 3A), except for the fact that during reversal learning, the food was placed in the arm that was opposite to the one that was rewarded during training. In the fi rst reversal session, both sedentary and exercise animals 97

99 Chapter 7 # DC Xp - ositive ce lls % D CX-p ositive d en d rites naive acquisition * naive acquisition Sedentary control * Enhanced neurogenesis A B Figure 5: Eff ect of acquisition on hippocampal neurogenesis. A) Y-maze acquisition caused a moderate reduction in immature neuron number in sedentary mice, but not in animals with enhanced neurogenesis. B) The density of DCXpositive dendrites in the granule cell layer was reduced in sedentary mice after acquisition of the Y-maze, but remained unchanged in mice with enhanced neurogenesis. C) Acquisition of the Y-maze did not cause changes in the number of pcreb-positive cells, neither in sedentary animals, nor in mice with enhanced neurogenesis. *: P Cell numbers are shown as the average number of cells/section ± S.E.M Sedentary control Enhanced neurogenesis #PCREB- positive ce lls naive acquisition C Sedentary control Enhanced neurogenesis showed an impaired performance, which was signifi cantly improved during the following sessions (Fig. 3C, P 0.001). The significant session*group interaction (F(7,105)=2.1, P 0.05) showed that runners learned to fi nd to new location of the food faster than sedentary mice. Y-maze retention and reversal learning reduce hippocampal neurogenesis Since we observed that enhanced neurogenesis facilitated acquisition and retrieval of spatial information, we further studied the effects of these aspects of Y-maze learning on hippocampal neurogenesis. The brain material was stained for the immature neuron marker DCX and for ser133-phosphorylated CREB (Fig. 4, page 142). DCX can be used as a suitable marker for investigating the absolute number of newly formed neurons (Rao and 98

100 Memory retrieval reduces neurogenesis # DCX-po sitive ce lls % DCX- positive de ndrites naive retention *** naive retention Sedentary control *** *** Enhanced neurogenesis *** A B Figure 6: Eff ect of memory retention on hippocampal neurogenesis. A) The retention test caused a signifi cant reduction in DCX-positive cell number, both in sedentary mice and animals with enhanced neurogenesis. B) The reduction is DCX-positive cell number was refl ected in the decline in DCX-positive dendrite density. C) Quantifi cation of pcreb-positive cells showed that memory retention also resulted in a decline in CREB phosphorylated cell number. *: P 0.05; ***: P Cell numbers are shown as the average number of cells/section ± S.E.M. Note that the data presented for the sedentary naive animals and the naive mice with enhanced neurogenesis are similar to Fig Sedentary control Enhanced neurogenesis # PCR EB- positiv e cells naive retention *** C 150 * Sedentary control Enhanced neurogenesis Shetty, 2004). CREB phosphorylation has also been reported to occur in newly formed, PSA-NCAM positive granule neurons during maturation (Nakagawa et al., 2002; Zhu et al., 2004b). Since there is an almost complete overlap between DCX-expression and NCAM polysialylation (Nacher et al., 2001a), we supposed that the pcreb-positive cells in the granule cell layer were in fact newly formed neurons. This assumption was confi rmed by the significant correlation that we found between the number of DCX-expressing neurons and the number of pcreb-positive cells (R 2 =0.62, P 0.001). In addition, a double labeling for DCX and pcreb was performed (Fig. 4, page 142). Random investigation of sections from the different experimental groups showed that the majority of the DCX-positive cells colocalized with pcreb. Occasionally, a DCX single labeled cell could be found. However, not all pcreb positive cells also expressed DCX, suggesting that some existing mature granule neurons contain pcreb. 99

101 Chapter 7 In order to statistically analyze potential effects of exercise or any of the Y-maze conditions on the number of newly formed hippocampal neurons, a two-way ANOVA was performed with the 4 different Y-maze groups (naive, acquisition, retention and reversal) and the housing condition (sedentary or exercise) as between-subjects variables. Both for DCX (F(1,66)=36.8, P 0.001) and pcreb (F(1,66)=36.4, P 0.001), a significant effect of exercise was observed, with runners having more DCX and pcreb-positive cells. The Y-maze groups also differed significantly from each other with respect to DCX-positive cell number (F(3,66)=20.7, P 0.001), DCX-positive dendrite density (F(3,66)=20.8, P 0.001) and pcreb-positive cell number (F(3,66)=12.1, P 0.001). For pcreb, there was also a near-significant interaction between exercise and learning (P 0.10). A post hoc LSD test was performed to test which Y-maze groups differed from naive controls. Y-maze acquisition did not cause major changes in newly formed cell numbers. A post hoc LSD-test showed that in the sedentary animals, a mild reduction was observed in DCX-positive cell number (Fig. 5A, P 0.05) and in the density of DCX-positive dendrites (Fig. 5B, P 0.05). However, the number of pcreb-positive neurons did not differ between Y-maze naive and trained mice (Fig. 5C). In mice with enhanced neurogenesis, Y-maze acquisition did not cause any changes in DCX-positive cell number (Fig. 5A), DCX-positive dendrites (Fig. 5B) or CREB phosphorylation in the granule cell layer (Fig. 5C). Next, we studied whether re-exposure to the Y-maze task 14 days after mice had acquired the task caused changes in the number of newly formed cells. Figure 6A shows a significant reduction in DCX-positive cell number after memory retention, both in sedentary mice (P 0.001) and animals with enhanced neurogenesis (P 0.001). The DCXpositive dendrite measurements also revealed a decline in animals that were exposed to the retention test (Fig. 6B, sedentary: P 0.001, enhanced neurogenesis: P 0.001). In addition, the number of pcreb-positive cells was significantly reduced after memory retention (Fig. 6C, sedentary: P 0.05), enhanced neurogenesis: P 0.001). Furthermore, we investigated hippocampal neurogenesis after relocation of the food reward, 14 days after mice had acquired the Y-maze task. Reversal learning resulted in changes in the number of newly generated neurons that were comparable to those observed after memory retention. Both DCX-positive cell number (Fig. 7A, sedentary: P 0.001, enhanced neurogenesis: P 0.001) and dendrite density (Fig. 7B, sedentary: P 0.001, enhanced neurogenesis: P 0.001) were significantly reduced after reversal learning. In addition, we found a significant decrease in the number of pcreb-positive cells after reversal learning (Fig. 7C, sedentary: P 0.05, enhanced neurogenesis: P 0.001). All changes reported for DCX-immunoreactivity or pcreb-positive cell number were visible in both blades (separate data set not shown). 100

102 Memory retrieval reduces neurogenesis # DCX-positive cells % DC X -positive dendrites naive reversal naive reversal *** Sedentary control *** *** Enhanced neurogenesis *** A B Figure 7: Eff ect of reversal learning on hippocampal neurogenesis. A) Both in sedentary animals and in mice with enhanced neurogenesis, reversal learning caused a decrease in the number of immature neurons. B) Also the coverage of the granule cell layer with DCX-positive dendrites was reduced upon reversal learning. C) The number of pcreb-positive cells was signifi cantly lower in animals that were exposed to the reversal learning paradigm. *: P 0.05; ***: P Cell numbers are shown as the average number of cells/section ± S.E.M. Note that the data presented for the sedentary naive animals and the naive mice with enhanced neurogenesis are similar to Fig Sedentary control Enhanced neurogenesis #PC REB-positive cells naive reversal *** C 150 * Sedentary control Enhanced neurogenesis 101

103 Chapter 7 Discussion Benefi cial eff ects of exercise on Y-maze acquisition, retention and reversal learning Our data show that 14 days of exercise increased the rate of acquisition in the Y-maze, improved retention of previously acquired information and facilitated reversal learning. The fact that exercise had a positive effect on Y-maze acquisition is in line with other studies (Fordyce and Farrar, 1991; Fordyce and Wehner, 1993; Shaw et al., 2003; Van Praag et al., 1999a) and supports the idea that newly formed neurons participate in memory formation (Gould et al., 1999a; Snyder et al., 2005). Newborn neurons form a specifi c population of cells that may serve as a substrate for the formation of new memories. They are not yet involved in other memory traces and they show extraordinary morphological (Hastings and Gould, 1999; Seki and Arai, 1991; Seki and Arai, 1993) and synaptic (Schmidt-Hieber et al., 2004; Snyder et al., 2001) plasticity. They may therefore play a specific role in the formation of memories (Gould et al., 1999c). Here we present that physical exercise not only promotes the acquisition of a spatial learning task, but that 14 days of running wheel activity, starting after mice have mastered a task, also has benefi cial effects on retention of information and on the reversal learning ability. These data contradict with our initial hypothesis, based on the report by Feng and colleagues (2003), stating that hippocampal neurogenesis serves to remove old data from the hippocampus. However, there are major differences in the experimental setup between the present study and the experiments by Feng et al. We used another learning task (Y-maze versus contextual fear conditioning), a different method to increase neurogenesis (wheel running instead of environmental enrichment) and wildtype mice instead of knockout animals. Nevertheless, with our experimental paradigm we cannot confi rm that hippocampal neurogenesis erases existing memory traces from the hippocampus. The question remains whether and how the elevated levels of hippocampal neurogenesis may have contributed to the improved memory retrieval. The excess in newly generated cell number was created after the animals had mastered the task. This implies that the extra neurons that were produced during exercise were not involved in memory formation. However, the enhanced hippocampal neurogenesis following the acquisition may have caused a more effi cient memory consolidation. An increased capacity of the granule cell layer reduces the chance that the cells that were involved in Y-maze acquisition are subsequently involved in other memory traces. The incorporation of a cell into multiple memory traces increases the likelihood for mistakes during retrieval (Deisseroth et al., 2004). Increasing the capacity of the granule cell layer may minimize this problem. The positive effects of physical exercise on various aspects of Y-maze learning can probably not all be attributed to the increase in hippocampal neurogenesis. Directly after 14 days of exercise, the rise in pcreb-positive cell number exceeded the increase in the number of DCX-positive cells, suggesting that the improved learning performance could have been caused by increased transcription of genes that promote long-term memory formation (Bourtchuladze et al., 1994; Guzowski and McGaugh, 1997; Kida et al., 2002; 102

104 Memory retrieval reduces neurogenesis Kogan et al., 1997; Silva et al., 1998). In addition, running has been shown to increase levels of various growth factors and neurotrophic factors (Fabel et al., 2003; Farmer et al., 2004; Gomez-Pinilla et al., 1997; Oliff et al., 1998; Trejo et al., 2001) and to promote, for instance, cerebral blood flow (Endres et al., 2003), angiogenesis (Swain et al., 2003) and cholinergic synaptic communication (Fordyce and Farrar, 1991), all of which may positively influence learning and memory. Finally, it is important to realize that exercise-induced improvements in learning, retention and reversal learning are not necessarily mediated by the same mechanism. Memory retention and reversal learning reduce hippocampal neurogenesis both in sedentary mice and animals with enhanced neurogenesis We further determined the effects of training, memory retention and reversal learning in the Y-maze on the number of newly formed neurons. Hippocampal neurogenesis was determined by staining for DCX and pcreb. DCX is expressed by immature neurons (Brown et al., 2003b; Couillard-Despres et al., 2005) and quantifi cation of DCX expression therefore provides direct information on the number of newly formed neurons (Rao and Shetty, 2004). CREB phosphorylation has also been shown to occur in maturing neurons (Fujioka et al., 2004; Nakagawa et al., 2002). We found a highly signifi cant correlation between pcreb-positive cell number and DCX-expression. Moreover, double labelings confi rmed that DCX and pcreb were strongly colocalized in the granule cell layer and therefore both predominantly represent immature neurons. We show that training in the Y-maze did not cause changes in immature neuron number in mice with enhanced neurogenesis. These data confi rm the results of other studies, also from our own group, in which spatial learning had no effect on hippocampal neurogenesis (Snyder et al., 2005; Van der Borght et al., 2005c; Van Praag et al., 1999b). However, in sedentary mice, Y-maze acquisition resulted in a reduction in DCX-expressing cells, but not in pcreb-immunoreactive cells. Hippocampal neurogenesis has been shown to be differentially regulated throughout the different phases of learning (Dobrossy et al., 2003). The fact that we found changes in DCX-expression in sedentary animals, but not in mice with enhanced neurogenesis may be due to the fact that the two groups were in a different phase of the learning curve at the moment of termination. Sedentary mice had just reached the level of 80% correct trials, whereas the mice with enhanced neurogenesis already performed at maximal levels for 3 days. We further explored the effects of memory retention and reversal learning on hippocampal neurogenesis. Both paradigms caused a dramatic reduction in the number of newly formed cells, both in control animals and in mice with enhanced neurogenesis. The common feature of memory retention and reversal learning is that both processes require retrieval of the stored spatial information, so it can be suggested that retrieval of spatial information reduced hippocampal neurogenesis. The decrease in the number of newly formed neurons can be caused by increased apoptosis of newly formed cells. Alternatively, the generation of new cells may have been inhibited during retention testing and reversal learning. A third option is that memory 103

105 Chapter 7 retrieval caused an accelerated maturation of neural precursors. However, since the length of DCX-expression has been shown to remain stable, even in conditions such as aging or hippocampal injury (Rao and Shetty, 2004), the latter explanation is not very likely. The data we present here suggest that neurogenesis in the dentate gyrus is actively suppressed during memory retrieval. Other studies have shown a reduction in hippocampal activity after memory retrieval. Retention testing of mice in a radial maze, 25 days after acquisition, decreased hippocampal metabolic activity, measured by ( 14 C)-deoxyglucose uptake, below baseline levels (Bontempi et al., 1999). In addition, repeated exposure of rats to a familiar environment significantly reduced hippocampal CREB phosphorylation compared to naive controls (Winograd and Viola, 2004). However, in those studies no differentiation was made between the different hippocampal subregions. The changes we observed occurred in the dentate gyrus, a part of the hippocampus which is known to be crucially involved in the formation of memories, but much less in the retrieval of information (Eldridge et al., 2005; Lee and Kesner, 2004). The fact that neurogenesis in the dentate gyrus is inhibited during retrieval, suggests that the presence of highly plastic newly formed neurons in the gateway to the hippocampus may negatively affect the retrieved memory. Fourteen days after the animals have learned the Y-maze, the acquired information is most likely stored in the cortex and retrieval is independent of the hippocampus (Beylin et al., 2001; Kim and Fanselow, 1992; Takehara et al., 2003). However, re-exposure to the same context is thought to re-activate the hippocampal memory trace and to return it to a more labile state which is sensitive to disruption (Debiec et al., 2002; Nader, 2003). We hypothesize that during this labile state of the retrieved memory, neurogenesis is inhibited in order to prevent the formation of a new, but redundant, memory trace of a familiar context. We propose that the simultaneous retrieval and formation of a memory of the same contextual situation may cause interference and hinder the consolidation process. In summary, we show here that 14 days of physical exercise facilitates acquisition, memory retention and reversal learning in the Y-maze. Possibly, these benefi cial effects on various aspects of spatial learning are mediated by the increase in hippocampal neurogenesis after exercise, which persisted for at least six days after the running wheel had been removed from the cage. Secondly, we show a signifi cant reduction in hippocampal neurogenesis following retrieval of spatial information in the Y-maze. Since newborn neurons may form an important substrate for the formation of new memory traces, we propose that active suppression of neurogenesis during re-exposure to a familiar environment may prevent the formation of redundant memories and reduce possible interference between the existing and the newly formed memory trace. 104

106 Memory retrieval reduces neurogenesis Acknowledgements We thank Jan N. Keijser for helping us with the DCX-measurements. This work was supported by the Dutch Organization for Scientific Research (NWO-vernieuwingsimpuls to E.A.V.d.Z. ( )). 105

107 106

108 8 Chapter General Discussion 107

109 Chapter 8 CONTENTS 1. Summary of the results 2. Methodological considerations 2.1. Proliferation of hippocampal progenitors: functional interpretation 2.2. Differentiation and survival: BrdU versus doublecortin 3. Neurogenesis and the medial septum 3.1. Regulation of cell proliferation by the medial septum 3.2. Regulation of cell survival by the medial septum 3.3. Functional implications of the regulatory role of the medial septum in hippocampal neurogenesis 4. Neurogenesis and exercise 4.1. General remarks 4.2. Potential mechanisms causing the exercise-induced increase in neurogenesis 4.3. Is the increase in hippocampal neurogenesis during running functional? 5. Neurogenesis and learning 5.1. Arguments supporting a role for neurogenesis in learning Newly formed neurons show strong synaptic plasticity Newly formed neurons show strong morphological plasticity Newly formed neurons facilitate memory consolidation Memory retrieval reduces neurogenesis 5.2. Model on how new neurons are implicated in learning and memory consolidation Potential mechanisms underlying the decrease in neurogenesis during memory retrieval Can our model explain supposed discrepancies in the literature? Does neurogenesis have a function in humans? 6. Recommendations for future research 7. Concluding remarks 108

110 General Discussion 1. Summary of the results The objective of this thesis was to gain more insight into the potential function of newly formed neurons in the adult hippocampus, particularly in learning and memory processes. In order to study this, the effects of behavioral, physiological and neurobiological manipulations on hippocampal neurogenesis were explored. In chapter 2 we determined basal levels of hippocampal cell proliferation across the day. In contrast to what might be expected on the basis of fi ndings in other tissues and on the well-known daily rhythms in behavioral activity and physiological parameters in the hippocampus, we found that cell proliferation in the dentate gyrus (DG) takes place at a constant rate throughout the day. We further examined whether disturbance of the normal sleeping behavior of the animal would influence the steady rate in hippocampal cell proliferation. Our data show that 12 h of sleep deprivation during the resting phase of the mice did not affect the number of hippocampal dividing cells. Next, we studied whether running wheel activity, a condition known to enhance hippocampal neurogenesis, causes an elevation of the basal steady rate in cell proliferation or whether the effects are most profound directly following the active dark phase. We observed an increase in the number of proliferating cells immediately following the period of highest activity. At the end of the resting phase, the number of proliferating cells in runners did not significantly differ anymore from sedentary controls. This study shows that the basal steady rate of hippocampal cell proliferation can be changed by behavioral interventions, indicating that it is not a prerequisite for hippocampal functioning to have a constant level of proliferating cells. In chapter 3 we further explored the effects of enhanced physical exercise on hippocampal neurogenesis. We investigated whether effects of exercise on neurogenesis are acute or require chronic running wheel activity. First, hippocampal cell proliferation and the formation of new neurons were determined after 1, 3 and 10 days of exercise. Next, we investigated the dynamics in neurogenesis when, after 10 days of exercise, the wheel was removed from the cage. Cell proliferation and immature neuron number were only significantly increased after 10 days of exercise. Furthermore, we observed that 24 h after removal of the wheel, cell proliferation was normalized again, suggesting that the number of proliferating progenitors is acutely regulated by physical activity. However, even 6 days after withdrawal of the running wheel, the number of immature neurons was still elevated, indicating that the survival promoting effects of exercise are persisting. This study shows that hippocampal cell proliferation and cell survival are mediated by independent mechanisms. Chapter 4 describes a study in which the role of the input from the medial septum (MS) on hippocampal neurogenesis was investigated. NMDA was injected into the MS of rats, in order to induce excitoxicity. Cholinergic cell number in the MS was only decreased after infusion of 60 nmol of NMDA. However, both 30 and 60 nmol of NMDA caused a strong reduction in the number of GABAergic neurons. Hippocampal cell proliferation in lesioned animals did not differ from sham rats, but the lesion resulted in a significant re- 109

111 Chapter 8 duction in the number of cells that had been labeled with BrdU 1 week prior to the lesion. These data suggest that the GABAergic component from the septohippocampal pathway may be involved in the regulation of survival of newly formed cells. In chapter 5 rats were exposed to either 1 or 4 days of active shock avoidance (ASA) learning. ASA is a hippocampus-independent associative learning task, but the DG has been shown to be activated upon ASA learning. We show that neither cell proliferation, nor the survival of newly formed cells was affected by any of the ASA paradigms. We also determined plasma corticosterone levels throughout the learning procedure. There were no differences between animals that had received foot shocks in the shuttle box and subjects that were only habituated to the box, indicating that potential effects of learning on neurogenesis were not masked by stress-induced changes in the number of newly formed cells. From this study we can conclude that this form of hippocampus-independent associative learning does not influence the number of newly formed granule neurons. The experiments in chapter 6 were performed to test whether training in a hippocampus-dependent learning task, the Morris water maze (MWM), induces changes in hippocampal neurogenesis in two rats strains, Wistar and Sprague-Dawley. We did not observe any changes in cell proliferation or newly formed cell survival after spatial learning. However, we found an increased number of PSA-NCAM positive neurons in the granule cell layer of animals that had been exposed to the spatial learning task, indicating increased morphological plasticity in this brain area. Furthermore, we showed a significant difference in newly formed cell survival between the rat strains that were investigated, with Wistars having significantly more BrdU-positive cells than Sprague-Dawleys. This study shows that, despite plastic changes occured in the DG upon MWM learning, the rate of hippocampal neurogenesis remained unchanged. In chapter 7, mice were housed with a running wheel in order to enhance neurogenesis. The effects of the increased number of newly formed neurons on Y-maze acquisition and retrieval were investigated. We show that mice with enhanced neurogenesis faster acquired the task, showed improved memory retention and performed better in the reversal learning test. Next, we examined the effects of retrieval of spatial information on the number of newly formed neurons. We showed that re-exposure to the Y-maze 14 days after acquisition resulted in a signifi cant reduction in the number of newly formed neurons. These data indicate that elevated levels of hippocampal neurogenesis have benefi cial effects on learning and memory. However, memory retrieval causes a reduction in the number of immature neurons, perhaps to reduce the chance of interference between retrieved information and newly formed memory traces. 110

112 General Discussion 2. Methodological considerations In order to determine changes in the number of newly formed neurons, different aspects of hippocampal neurogenesis can be studied. The total cell number that is newly formed depends on the rate of proliferation of undifferentiated progenitors, the percentage of these cells that differentiate into a neuronal phenotype and the number of newly generated neurons that survive and become functionally incorporated into the hippocampal circuitry. Here, we discuss that hippocampal cell proliferation might not be the most relevant aspect of the neurogenic process to investigate when studying the regulation of neurogenesis and the potential consequences for hippocampal functioning. Survival of newly formed cells and the phenotype that they acquired might be much more important parameters to investigate. Secondly, we compare two different methods to investigate differentiation and survival of newly formed cells Proliferation of hippocampal progenitors: functional interpretation Hippocampal cell proliferation has been reported to rapidly respond to a large variety of stimuli, such as growth factors, steroid hormones, stress or physical activity. The functional interpretation of these rapid changes in proliferating cell number is diffi cult. An increase in the generation of new cells does not directly result in an increase in the number of new neurons. It takes a couple of weeks before progenitors have fully differentiated into mature and functional neurons. Therefore, changes in proliferating cell number do not have direct implications for hippocampal functioning, but will only potentially affect the number of functional newly formed neurons over time. In addition, an acute change in proliferation rate might be normalized readily after the stimulus that caused the change in proliferating cell number has disappeared. Therefore, a single observation of alterations in the number of dividing progenitors does not allow conclusions on the number of new neurons that will be formed. Acute stress, for instance, has been shown to cause an acute reduction in hippocampal cell proliferation, but proliferation rate is normalized again within 24 h (Heine et al., 2004b). Acute stress will therefore not have a major impact on the total number of neurons that will be generated. A different example, illustrating that based on a single measurement of hippocampal cell proliferation incorrect conclusions on neurogenic rate may be drawn, is the effect of estradiol on cell proliferation. Four hours after a single injection with estradiol, cell proliferation in female rats is increased. However, 48 h after the estradiol injection, a reduction in cell proliferation can be observed (Ormerod et al., 2003). The net rate of hippocampal cell proliferation does not differ between males and females (Perfi lieva et al., 2001; Tanapat et al., 1999). 111

113 Chapter 8 The data reported in chapter 2 also indicate that fluctuations in cell proliferation may not have direct consequences for the animal. We demonstrated that hippocampal cell proliferation takes place at a stable rate across the day, but that enhanced physical exercise caused a peak in proliferation directly after the period of highest activity, which is normalized again at the end of the resting phase. If it was important for the hippocampus to maintain a stable rate in hippocampal cell proliferation, exercise would have resulted in a constant elevation in proliferating cell number at all times of day. Therefore, our data suggest that acute changes in the number of proliferating cells will not have direct implications for hippocampal functioning. Alterations in the number of proliferating cells may only have functional consequences if the effect is persisting. When the production of new cells is changed for a prolonged period of time, this might affect the net number of new neurons that will be generated. However, chronic changes in hippocampal cell proliferation will only result in changes in the number of newly formed neurons if the percentage of apoptotic cells does not change concomitantly with the proliferating cell number. Therefore, even chronic changes in hippocampal cell proliferation do not allow direct conclusions about the potential impact on in hippocampal functioning, unless it can be shown that the number of newly formed neurons or glia cells has signifi cantly changed. In conclusion, we can say that the upregulation or downregulation of hippocampal cell proliferation is not inevitably refl ected by an increase or decrease in the number of new neurons that are generated. Moreover, a change in proliferating cell number will only affect the number of newly formed cells over time. Many of the newly generated neurons die within 1 to 4 weeks after they were formed (Dayer et al., 2003; Hastings and Gould, 1999), indicating that there is always a surplus of neurons generated. If a change in newly formed cell number is required, it would be much more effi cient to adjust the percentage of apoptotic cells Differentiation and survival: BrdU versus doublecortin In order to determine changes in hippocampal neurogenesis, the number of newly formed cells that survive and the phenotype that these cells acquired are the most relevant parameters to investigate. The two most commonly used methods to study the number of newly formed hippocampal neurons were also applied in this thesis. First of all, animals can be injected with BrdU and sacrificed a few weeks later. A triple labeling with a glial and a neuronal marker can be performed to study the phenotype of the newly formed cells. A second method is to stain brains for doublecortin (DCX), a protein that is associated with the cytoskeleton and is almost exclusively expressed in immature neurons (Brown et al., 2003b; Couillard-Despres et al., 2005; Rao and Shetty, 2004). Both methods have advantages and disadvantages with respect to the interpretation of the results. The temporal resolution of BrdU labeling is relatively high, because BrdU only labels cells within two hours after injection. By varying the timing of the BrdU injection 112

114 General Discussion relative to the experimental treatment, the age of cells that are investigated can be precisely determined. DCX is expressed in maturing neurons for a few days (mice) or even weeks (rats) and does therefore not provide exact information on the age of the stained cells. The fact that BrdU only labels cells in a narrow time window is also one of the disadvantages of this method. The timing of the injection relative to the experimental handling is of crucial importance for the outcome of the experiment. Possible changes in newly formed cell number could easily be missed because cells were labeled at an irrelevant time point compared to the experimental treatment. DCX is expressed in cells for days or weeks, which increases the chance of observing changes in newly formed cell number. Another diffi culty with BrdU labeling is that there is a risk for label dilution. When a BrdU labeled cell divides, the daughter cells are also BrdU-positive, but the label is diluted. Therefore, the number of BrdU-positive cells that is detected at a certain time point after injection is a result of the proliferation rate of the cells, the apoptotic rate and the dilution of the label. At a certain time point after labeling with BrdU (approximately 3 days in mice and 7 days in rats), a steep decline in labeled cell number can be observed. This reduction may not be exclusively caused by apoptosis. Cells could have continued to proliferate, but as they divide every 24 h (Cameron and McKay, 2001), the label can theoretically be diluted 27 = 128 times (Prickaerts et al., 2004). Dayer et al. (2003) showed that BrdU-positive cells gradually lose Ki-67 immunoreactivity between 1 and 4 days after labeling. Also from this study it cannot unequivocally be concluded that BrdU labeled cells became postmitotic. Perhaps, the BrdU signal of cells that continue to divide for 4 days becomes so much diluted that it is not visible anymore. The BrdU-positive cells that are counted a few days after labeling may only be cells that quickly exited the cell cycle after the labeling. However, the fact that many TUNEL-positive cells can be observed in the subgranular zone (Biebl et al., 2000) suggests that there is a considerable level of apoptosis in the newly formed cell population and that the decline in labeled cell number most likely reflects a combination of apoptosis and label dilution. DCX as a marker for newly formed neurons also has a few disadvantages. A part of the DCX-positive cell population is still mitotic (Kempermann et al., 2004). DCX-immunostaining can therefore not differentiate between cell proliferation and cell survival. Furthermore, changes in DCX-immunoreactive cell numbers may not only reflect changes in generation or death of newly formed neurons, but might also indicate that the maturation of cells is delayed or accelerated without influencing the net number of functional new neurons that is generated. However, the time window in which DCX is expressed appears to be very stable. Even in rather severe conditions, such as hippocampal damage, the length of DCX-expression remains unaltered (Rao and Shetty, 2004). DCX has been reported to be exclusively expressed by immature neurons (Couillard-Despres et al., 2005). However, in our studies, we also observed DCX-positive cells in non-neurogenic regions, such as the piriform cortex and some hypothalamic subregions (unpublished observation). This indicates that DCX might also be expressed by cells that undergo, for instance, morphological reorganization. Quantifi cation of the number of DCX-positive cells may therefore overestimate the number of newly formed neurons. 113

115 Chapter 8 However, careful comparisons between BrdU labeling and DCX analysis have shown that DCX is a reliable marker for the number of newly generated neurons in the hippocampus (Rao and Shetty, 2004). Since BrdU labels all proliferating cells, double or triple labelings may provide information on the complete distribution of phenotypes of the newly formed cells. DCX-expression is restricted to newly formed neurons and therefore, changes in gliogenesis or the formation of endothelial cells, for instance, will not be detected. However, an advantage of DCX-analysis is that it also provides information on the morphological maturation of the newly formed neurons, since it clearly stains the dendritic tree of the cells. Finally, BrdU is a carcinogenic substance that has been shown to induce malformations in the developing brain (Kolb et al., 1999). Therefore, there is a potential risk that BrdU labeled neurons in the adult hippocampus do not function similarly to non-labeled neurons. In summary, it can be concluded that BrdU labeling as well as DCX staining are suitable methods to detect newly formed neurons, but that both methods have advantages and disadvantages. Perhaps, for every experimental situation, it should be decided which method is most suitable for answering the main question. 3. Neurogenesis and the medial septum 3.1. Regulation of cell proliferation by the medial septum The data reported in chapter 4 show that infusion of 60 nmol, but not 30 nmol of NMDA into the MS caused a significant reduction in the number of cholinergic neurons, but no change in the number of proliferating cells in the hippocampus. However, we found a significant difference in cell proliferation between the two lesioned groups, with the rats that were injected with 60 nmol of NMDA having fewer proliferating cells than animals that received the lower dose of NMDA. Interestingly, measurements of cholinergic fi ber density in the DG showed that animals that were lesioned with 30 nmol of NMDA had significantly more ChAT-positive fi bers in the molecular layer of the DG than rats that were infused with 60 nmol of NMDA (Fig. 1). These data suggest that 30 nmol of NMDA may have activated, instead of damaged, the MS, and that the difference in proliferating cell number between the two lesioned groups might be related to the difference in cholinergic input to the hippocampus. This hypothesis is supported by data we obtained 4 weeks after an electrolytic lesion of the MS in mice (Fig. 2), which reduced cholinergic input to the hippocampus by 50% (Fig. 3). The lesion resulted in a significant decline in the number of proliferating cells in the hippocampus (Fig. 4A-C). There is more evidence for a role of acetylcholine in the regulation of progenitor proliferation. Almost complete ablation of hippocampal acetylcholine levels by infusion of 192IgG-saporin into the lateral ventricles resulted in a mild, but significant reduction in hippocampal cell proliferation (Mohapel et al., 2005). 114

116 General Discussion ChAT-positive fiber density 40 # * sham 30 nmol 60 nmol Figure 1: Cholinergic fi ber density in the hippocampus after lesion of the MS. Thirty or 60 nmol of NMDA was infused into the MS of Wistar rats. Animals were sacrifi ced one week after the lesion and brains were stained for the acetylcholine synthesizing enzyme choline acetyl transferase (ChAT). The area of the molecular layer of the DG that was covered with ChAT-immunoreactive fi bers was determined. The graph shows that cholinergic fi bers density in rats that were lesioned with 30 nmol of NMDA was signifi cantly higher than after infusion of the highest NMDA concentration. In addition, 30 nmol of NMDA caused a near-signifi cant increase in cholinergic fi ber density compared to sham rats. #: P 0.10, *: P Sham: n=7, 30 nmol of NMDA: n=6, 60 nmol of NMDA: n=5. In parallel, repeated systemic injections with the muscarinic receptor agonist oxotremorine (Fig. 5) or with the acetylcholine esterase inhibitor physostigmine (Mohapel et al., 2005) caused a moderate increase in the number of dividing hippocampal progenitors in mice and rats, respectively. Taken together, it can be suggested that the cholinergic component of the septohipocampal pathway may affect hippocampal cell proliferation, but only after severe manipulations Regulation of cell survival by the MS Our data indicate that the GABAergic component of the septohippocampal pathway might be involved in the regulation of survival of newly generated cells. Infusion of NMDA into the MS resulted in a significant reduction of hippocampal neurons and glia cells that had been generated 7 to 9 days prior to the lesion. The lesion with 30 nmol and with 60 nmol of NMDA caused a similar decrease in cell survival. Cholinergic cell number in the MS was only decreased after the 60 nmol lesion, but the number of GABAergic neurons was strongly reduced after infusion of both concentrations of NMDA, suggesting that the loss 115

117 Chapter 8 A B C D E F Figure 2: Representative electrolytic lesion in a mouse. A unipolar macroelectrode was inserted into the MS of C57Bl/6 mice and an electric current of 1 ma was applied for 12 s. Animals were sacrifi ced 29 days after surgery and brains were stained for GFAP in order to determine the extent and the position of the lesion. A to E show successive sections of the MS, with an intersection interval of 60 mm. The white arrows point at the position of the core of the lesion. The grey area surrounding the core contains many GFAP-positive astrocytes, which refl ects the penumbra of the lesion. F shows a magnifi cation of the selected area in B, in which the reactive astrocytes are clearly visible. A-E: scale bar = 200 μm, F: scale bar = 50 μm. of GABAergic neurons caused the reduction of hippocampal neurogenesis that we observed. Also after an electrolytic lesion of the MS in mice, a significant reduction in the number of newly formed neurons was observed (Fig. 4D-F). The decline in the number of immature neurons was stronger than the reduction in proliferation rate, suggesting that the lesion also affected cell survival. Although we do not have exact data on the extent of GABAergic neuron loss after the electrolytic lesion, the size of the lesion (Fig. 1) suggests 116

118 Are a % o f ChA T-p ositive fibers General Discussion B A GCL H Mol Mol *** *** *** Inner Outer Total C H Mol Mol GCL Sham Lesion Figure 3: ChAT-immunoreactivity in the DG after electrolytic lesion of the MS in mice. A) Both in the inner and the outer blade of the molecular layer, the percentage of the area covered with ChAT-positive fi bers was signifi cantly reduced by the lesion. A representative photomicrograph of ChAT-positive fi bers in an intact mouse is presented in B. Panel C shows cholinergic innervation of the DG in a lesioned animal. ***: P GCL: Granule cell layer, H: Hilus, Mol: Molecular layer. Scale bar = 75 μm. Sham: n=7, lesion: n=8. that the small current that we applied to the MS also caused significant damage to the GABAergic component of the MS. Our hypothesis that the GABAergic, and not the cholinergic, input to the hippocampus is involved in the regulation of cell survival is supported by data from the literature. Specific lesion of cholinergic neurons in the MS only caused a reduction in hippocampal cell proliferation, but not in survival of newly formed neurons (Mohapel et al., 2005). There are different mechanisms possible underlying the effects of GABAergic cell loss in the MS on cell survival in the hippocampus. GABA is known to exert non-synaptic effects during various steps of nervous system development (Owens and Kriegstein, 2002), such as synaptogenesis (Belhage et al., 1998), neuronal differentiation (Nguyen et al., 2003) and neurite extension (Wolff et al., 1978). GABA may therefore act as a kind of neurotrophic factor for immature or nearly mature hippocampal neurons that are formed during adulthood. Perhaps, the lesion caused a sudden fall in hippocampal GABA levels, which reduced neurotrophic support for the survival of newly generated cells in the DG. A different, and perhaps more likely, explanation would be that the reduced GABAergic input to the hippocampus resulted in a disturbed disinhibition. In intact animals, GAB- Aergic neurons in the MS project to hippocampal interneurons. This means that the inhibitory effect of hippocampal interneurons on hippocampal principal neurons is weakened by the GABAergic input from the MS. Damage to the GABAergic neurons in the MS would therefore result in an increased inhibition of hippocampal principal neurons. The downregulation of total hippocampal activity might underlie the decline in newly formed cell survival. There is experimental evidence suggesting a direct link between hippocampal activity and neurogenesis. Factors that are likely to increase hippocampal activity, such as environmental enrichment or physical exercise, promote survival of newly generated cells. It has also been reported that excitatory stimulation of hippocampal progenitors in vitro can 117

119 Chapter 8 # B rdu-positive c ells A ** * B GCL H C GCL H Area % DCX-positive dendr ites D Inner Outer Total *** *** *** E GCL H F GCL H 0 Inner Outer Total Sham Lesion Figure 4: Eff ects of electrolytic lesion of the medial spetum in mice on hippocampal neurogenesis. A) The lesion caused a signifi cant reduction in the number of proliferating cells, mainly in the inner blade of the subgranular layer. B) Representative photomicrograph of immunocytochemically stained BrdU-positive cells. The selected area is enlarged in panel C. D) The MS lesion resulted in a decreased number of immature neurons, both in the inner and the outer blade of the granule cell layer. E) Example of DCX-positive cells in an intact mouse. F) The number of newly formed neurons is clearly reduced after the lesion. *: P 0.05, **: P 0.01, ***: P GCL: Granule cell layer, H: Hilus. Scale bar = 50 μm. Sham: n=7, lesion: n=8. promote the number of newly formed neurons (Deisseroth et al., 2004). It could therefore be hypothesized that a chronic suppression of hippocampal activity, as might be the case after a lesion of the MS, cannot prevent newly generated cells from dying of apoptosis Functional implications of the regulatory role of the MS in hippocampal neurogenesis Since only severe disturbance of the MS caused (relatively mild) changes in hippocampal cell proliferation, it cannot be expected that normal fluctuations in activity in the MS will have noticeable effects on the number of proliferating progenitors. However, under 118

120 # BrdU -positive cells General Discussion Saline * Oxotremorine Figure 5: Eff ects of repeated injections with oxotremorine on hippocampal cell proliferation. C57Bl/6 mice received two i.p. injections per day for 3 days with 0.5 mg/kg of the muscarinic receptor agonist oxotremorine. In order to prevent peripheral side eff ects, oxotremorine was combined with methyl-scopolamine (1 mg/kg), a muscarinic receptor antagonist that cannot cross the blood-brain barrier. The interval between the two daily injections was 4 h. Together with every second oxotremorine injection mice were injected with 300 mg/kg BrdU on the 3 experimental days. The graph shows that repeated injections with oxotremorine signifi cantly increased the number of BrdU labeled cells. *: P Saline: n=10, oxotremorine: n=10. conditions where there is a loss of cholinergic neurons in the MS, such as during aging, the chronic reduction in cholinergic input to the hippocampus may be strong enough to significantly suppress hippocampal cell proliferation. The role of the MS in cell survival appears to be more pronounced. Modulation of GABAergic activity in the hippocampus might directly influence the number of newly generated neurons, which could affect memory formation (see section 5.2). In parallel with our lesion study, chronic activation of the MS will supposedly result in an upregulation of hippocampal neurogenesis. Wheel running increases hippocampal acetylcholine levels (Dudar et al., 1979; Nilsson et al., 1990) and it is known to induce theta activity in the hippocampus, caused by synchronous fi ring of cholinergic and GABAergic neurons in the MS (King et al., 1998; Teitelbaum et al., 1975). Running wheel activity indeed promotes hippocampal neurogenesis, but the role of the MS in the exerciseinduced upregulation of neurogenesis still needs to be investigated. A potential experiment that could be performed to answer this question is to lesion the MS and to subsequently investigate whether wheel running is still capable of increasing neurogenesis. If exercise has no effect on neurogenesis in lesioned animals, this would indicate that the MS is an essential mediator in the exercise-induced effects on hippocampal neurogenesis. If neurogenesis is still increased in runners with a damaged MS, this would mean that an intact MS is not a prerequisite for the effects of running on neurogenesis and that exercise may be a good method to (partly) rescue newly generated hippocampal neurons from the negative effects of cell loss in the MS. A potential problem when testing this hypothesis is that the input from the MS is known to be important in the motivation to display voluntary movements (Lawson and Bland, 1993; Oddie et al., 1996). This was confi rmed by a pilot experiment that we per- 119

121 # Ki-67 po sitive c ells Chapter sedentary exercise Figure 6: No change in cell proliferation after exercise in combination with dietary restriction. Mice were housed with a running wheel for 14 days and sacrifi ced during the fi rst half of the light period following the last night of running. Animals were food restricted, starting 1 day prior to the period of exercise. The mice were daily fed during the second half of the light phase and received suffi cient food to maintain their bodyweight at approximately 85% of what they originally weighed. Brains were stained for Ki-67 in order to determine proliferating cell number. In contrast to ad lib fed mice, hippocampal cell proliferation was not enhanced by exercise in food restricted mice. For details about the experimental procedure, see chapter 7. formed with mice that received an electrolytic lesion of the MS and were subsequently housed with a running wheel. After 3 days of wheel access, lesioned animals had run significantly less than sham mice (3.8 ± 1.6 km versus 13.2 ± 3.8 km, P 0.001). Since we showed that the exercise-induced increase in cell proliferation is correlated with the distance run by the mice (chapter 3), the reduction in spontaneous motor behavior after the MS lesion causes serious problems with the interpretation of neurogenesis data. 4. Neurogenesis and exercise 4.1. General remarks This thesis confi rms earlier reports showing a positive effect of running wheel activity on hippocampal neurogenesis. We further explored the temporal dynamics of the exercise-induced increase in hippocampal neurogenesis and the effects of enhanced neurogenesis on acquisition and retrieval of a spatial learning task. All exercise experiments were done with mice which, in contrast to rats, show a strong tendency to voluntarily run great distances in wheels. Throughout the experiments described in this thesis, we observed that the distance run by the animals per day gradually increased over the exercise period. This may reflect the fact that the mice developed their skill for running in a wheel. It may also indicate that running is a positive reinforcement, which stimulates the animals to increase this type of behavior. 120

122 General Discussion Running wheel activity was mainly restricted to the active dark phase of the lightdark cycle. Under food restricted conditions (chapter 7), mice not only ran during the dark phase, but also showed anticipatory running behavior, which means that they were active for approximately one hour preceding the time that they were fed. The dietary restriction also caused a significant increase in activity levels. Under ad lib food conditions, mice ran maximally 12 to 14 km per night, whereas some of the food restricted animals ran up to 25 km in one night. This is in line with other reports, showing the development of hyperactivity when food is not continuously available (Kas et al., 2003; Morse et al., 1995; Pirke et al., 1993). Interestingly, although the number of newly formed neurons in the DG was increased, we did not see an exercise-induced increase in cell proliferation in the food restricted mice (Fig. 6). Perhaps, under these extreme conditions, in which animals have low energy intake, but high energy expenditure, neurogenic processes may be regulated in a different way. However, this does not have direct implications for the interpretation of the results described in chapter 7, since exercise in food restricted mice still resulted in more newly generated neurons. In all exercise-experiments described in this thesis, sedentary control animals were housed in a standard cage without a wheel. Therefore, the effects of exercise on hippocampal neurogenesis might not be a direct consequence of an increased level of physical activity, but a result of the fact that the presence of a wheel in the cage is a form of cage enrichment. However, since we found a direct correlation between the distance that the mice ran during the last night and the number of Ki-67 positive cells (chapter 3), it is not likely that only the presence of a running wheel is suffi cient to enhance neurogenesis. Our data indicate that the generation of new hippocampal cells is only stimulated if the animals reach a certain threshold activity level Potential mechanisms causing the exercise-induced increase in neurogenesis The mechanisms underlying the exercise-induced increase in neurogenesis can be numerous, but there is evidence that elevated levels of BDNF (Berchtold et al., 2002; Garza et al., 2004), bfgf (Gomez-Pinilla et al., 1997), vegf (Fabel et al., 2003) or IGF-1 (Trejo et al., 2001) play a role. The binding of growth factors, especially vegf and IGF-1, to their receptors may induce transcription of genes that can directly affect hippocampal cell proliferation or cell survival, but it may also indirectly stimulate neurogenesis by promoting the generation of new capillaries (Lopez-Lopez et al., 2004), which may extend the vascular niche in which a large part of neurogenesis takes place (Palmer et al., 2000). Growth factors are not the only potential mediators in the effects of exercise on hippocampal neurogenesis. Physical activity also causes elevations in, for instance, endogenous opioids (Persson et al., 2004) or noradrenalin (Kulkarni et al., 2002), which may have a stimulatory effect on hippocampal progenitor proliferation. Moreover, running promotes serotonergic input to the hippocampus (Bequet et al., 2001; Gomez-Merino et al., 121

123 Chapter ; Wilson and Marsden, 1996) and increases hippocampal theta wave activity, evoked by cells in the MS (Teitelbaum et al., 1975), both of which might increase hippocampal neurogenesis (Brezun and Daszuta, 2000; Calza et al., 2003; Cooper-Kuhn et al., 2004; Malberg et al., 2000). Exercise also increases expression levels of certain subtypes of the NMDA receptor and other glutamate receptors (Farmer et al., 2004), some of which have been shown to be crucial for the neurogenesis promoting effects of exercise (Kitamura et al., 2003). Finally, physical exercise increases the total amount of cerebral blood fl ow (Ide and Secher, 2000; Swain et al., 2003), which is accompanied by an increase in extracellular glucose (Bequet et al., 2001) and other nutrients that can affect hippocampal cell production. The data described in chapters 2 and 3 suggest that hippocampal cell proliferation is very acutely regulated by physical exercise. The number of dividing cells is increased directly following a night of running wheel activity, but is not significantly elevated anymore 12 h later at the end of the resting phase. In line with this observation, we found that removal of the running wheel from the cage of animals that had been exposed to a wheel for 10 days normalizes hippocampal cell proliferation to baseline levels within a day. The acute response of cell proliferation on enhanced activity suggests that the effects of exercise on proliferation are not caused by persistent changes in the hippocampus, but that they are most likely mediated by factors that are immediately increased upon enhanced physical activity and that can directly promote cell proliferation. Although the number of proliferating cells is acutely returned to baseline when the period of enhanced physical activity has ended, the number of newly formed neurons remains elevated for at least 6 days. Since there are hardly any cells that still express DCX 6 day after they have been generated (Kempermann et al., 2004), the excess in DCX-positive cells 6 days after removal of the running wheel does most likely not represent the extra cells that were formed during exercise. Therefore, our data suggest that exercise stimulates survival of newly generated neurons and that this effect persists even if the animals do not have access to a wheel anymore. From these fi ndings it can be concluded that the exerciseinduced increase in cell proliferation and cell survival are mediated via different mechanisms Is the increase in hippocampal neurogenesis during running functional? The fi nding that something rather trivial as physical exercise can induce significant changes in a brain area that is not directly involved in motor behavior raises the question whether the relation between running and neurogenesis is functional. In natural circumstances, traveling great distances is usually associated with exploring large territories, which requires a good spatial memory and perhaps an increased hippocampal capacity. There is some evidence suggesting that wild-living animals with large territories that are frequently explored, have indeed more hippocampal neurogenesis than species which live in smaller 122

124 General Discussion areas (Amrein et al., 2004). Also in laboratory animals there is a potential relation between the distance that animals walk in their habitat and levels of neurogenesis. The enriched housing paradigm, which is known to enhance neurogenesis, includes housing in large cages and therefore an increase in physical activity. On the other hand, animals that are housed under standard laboratory conditions do not have much opportunity for physical exercise. Regarding the large distances mice run if they have access to a wheel, laboratory animals can be considered restricted in their endogenous drive to move. Perhaps, the point of view that exercise increases hippocampal neurogenesis should be converted to the statement that exercise-restriction suppresses the generation of new hippocampal neurons. Running may improve the hippocampal condition, which is suboptimal for neurogenesis under standard laboratory housing conditions. However, it should be noted that it is most likely not only enhanced physical activity that mediates the effects of exercise on neurogenesis. Mice that have access to a running wheel walk larger distances than wild living animals or animals in large cages. Wheel running not only is a form of increased activity, it may also be a rewarding type of behavior. For instance, voluntary wheel running in rats has been shown to cause an upregulation of the endogenous opioids dynorphin in the caudate putamen and met-enkaphalin in the hippocampus (Persson et al., 2004; Werme et al., 2000). Moreover, administration of the opioid receptor antagonist naltrexone partly blocked the exercise-induced enhancement of hippocampal cell proliferation (Persson et al., 2004), suggesting that the pleasure that animals experience during running also partly underlies the positive effect on neurogenesis. 5. Neurogenesis and learning 5.1. Arguments supporting a role for neurogenesis in learning One of the major aims of this thesis was to gain more insight into the potential link between neurogenesis and learning. At fi rst sight, the data that we obtained do not support the idea that hippocampal neurogenesis plays a role in memory formation. In none of the learning tasks that we studied (active shock avoidance, Morris water maze and Y-maze) acquisition resulted in an increase in the number of newly generated neurons. In addition, reports from the literature showing that animals with almost total ablation of hippocampal neurogenesis can still learn the Morris water maze or contextual fear conditioning (Madsen et al., 2003; Shors et al., 2002; Snyder et al., 2005), indicate that the presence of newly generated granule neurons is not a prerequisite for memory formation. Despite the fact that we did not fi nd training-induced increases in hippocampal neurogenesis and despite the fact that hippocampal neurogenesis does not appear to be a 123

125 Chapter 8 prerequisite for many learning tasks, we cannot exclude the possibility that newly formed neurons, if present, are incorporated into a memory trace. Below, we will provide some arguments that support a role for newly generated neurons in learning and memory processes Newly formed neurons show strong synaptic plasticity Granule neurons that are formed during adulthood show extraordinary synaptic plasticity. The young population of granule neurons has a lower current threshold in order to evoke an action potential and they can, in contrast to mature granule neurons, generate low-threshold Ca 2+ spikes under physiological conditions (Schmidt-Hieber et al., 2004). In addition, the induction of LTP is easier in granule neurons that are located close to the hilus compared to cells that can be found in the other parts of the granule cell layer (Schmidt-Hieber et al., 2004; Snyder et al., 2001; Wang et al., 2000). Taken together, these fi ndings indicate that newly formed granule neurons are more likely to become activated upon stimulation than mature ones, which makes them a likely substrate for memory formation Newly formed neurons show strong morphological plasticity Newly generated granule neurons are more susceptible to undergo plastic changes than existing cells. Morphological plasticity of newly formed neurons is demonstrated by the fact that they rapidly form axons towards the CA3 region. Newly formed neurons in the adult hippocampus have been shown to form dendritic growth cones during maturation, similar to embryonic development (Ribak et al., 2004). In rats, the formation of synaptic contacts between newly generated granule cells and CA3 neurons has been reported to occur already within 4 to 10 days after cells were formed (Hastings and Gould, 1999). Furthermore, maturing hippocampal neurons are positive for the plasticity markers DCX (Brown et al., 2003b; Couillard-Despres et al., 2005; Nacher et al., 2001a) and PSA-NCAM (Bonfanti et al., 1992; Kiss et al., 2001; Rousselot et al., 1995; Seki and Arai, 1993), both of which are involved in morphological reorganization. DCX is a microtubuleassociated protein, which is expressed by migrating and differentiating neuroblasts during embryonic cortical development (Francis et al., 1999). The role of DCX in microtubule polymerization suggests that DCX might also be involved in stabilization of microtubules and morphogenesis in differentiating neurons (Francis et al., 1999; Gdalyahu et al., 2004). Polysialylation of NCAM has also been associated with morphological plasticity of neurons. During development PSA-NCAM is abundantly expressed (Edelman, 1986; Seki and Rutishauser, 1998). The large PSA molecules are negatively charged and can thereby reduce adhesion forces between cells, allowing dynamic changes in membrane contacts, neurite outgrowth, neuronal migration, synaptogenesis and dendritic branching (Rutishauser et al., 1988). Moreover, PSA-NCAM can modify intracellular signaling 124

126 General Discussion cascades. It has been suggested that PSA-NCAM interferes with the interaction between the neurotrophic factor BDNF and its receptor TrkB, which can also induce morphological changes (Muller et al., 1996). Therefore, PSA-NCAM may be able to alter the morphology of hippocampal neurons either directly, by modulating cell adhesion, but also indirectly, via the BDNF signaling cascade. The hypothesis that the morphological plasticity of newly formed neurons increases the chance that they are involved in memory formation is supported by our fi ndings in chapter 6. There we showed that Morris water maze learning resulted in an increase in the number of PSA-NCAM immunoreactive cells in the granule cell layer. Since PSA-NCAM positive neurons were located in the deep part of the granule cell layer, close to the hilus, they most likely represent newly formed neurons undergoing morphological reorganization. Thus, the increase in PSA-NCAM positive cell number after Morris water maze learning may reflect spatial learning induced morphological alterations in the newly formed granule cell population, suggesting that these neurons were involved in memory formation Newly formed neurons facilitate memory formation Involvement of newly formed neurons in a memory trace is not necessarily reflected in an increased survival of this cell population. Learning may not affect the quantity of the newly generated granule cells, but it could have implications for the qualitative integration of those cells into the hippocampal network. Rats that had been exposed to cranial irradiation, which almost completely blocked hippocampal neurogenesis, performed as good as control animals during acquisition of the Morris water maze. Also, short-term memory was not affected by the irradiation procedure. However, in the long-term memory test, rats with reduced neurogenesis were significantly impaired (Snyder et al., 2005), suggesting that memory formation and consolidation in the absence of newly generated neurons is suboptimal. In the same line, treatment with cytostatic drugs during cage enrichment, prevented the enrichment-induced improvement in long-term, but not short-term, recognition memory (Bruel-Jungerman et al., 2005; Snyder et al., 2005). These data indicate that newly formed granule neurons can be part of a memory trace and facilitate memory formation and consolidation Memory retrieval reduces neurogenesis A fi nal argument supporting a role of newly formed granule neurons in learning and memory related processes, is our fi nding that retrieval of spatial information caused a reduction in immature neuron number in the granule cell layer. If neurogenesis had no function in learning at all, no changes in the number of immature neurons would have been expected. However, the fact that re-exposure to a familiar experimental set-up reduced hippocampal neurogenesis to a level under baseline indicates that newly generated neurons may somehow play a role in learning and memory. 125

127 Chapter Model on how newly formed neurons are involved in learning Combining the data from the literature and the results we described in this thesis, we here propose a model on how newly formed granule neurons could be involved in memory formation and retrieval (Fig. 7). Exposure of an animal to a certain stimulus, for instance a learning task or a novel environment, will activate granule cells in the DG, which can in turn stimulate a set of neurons in the CA3 region and indirectly the hippocampal output regions CA1 and subiculum. The set of hippocampal neurons that is activated upon exposure to a certain stimulus is named a memory trace. Because of their exceptional synaptic and morphological plasticity, we predict that activation of a newly formed neuron will result in stronger activation of the output area than activation of older granule neurons. Memory formation can still take place without neurogenesis, but we hypothesize that in the absence of newly generated granule neurons the memory trace will be weaker compared to the situation with neurogenesis. This hypothesis would explain the impaired long-term memory that has been observed in rats with reduced levels of neurogenesis (Bruel-Jungerman et al., 2005; Snyder et al., 2005). Impaired memory retention in the absence of neurogenesis may not only be caused by the fact that the memory trace is weaker. Especially when animals are exposed to multiple learning tasks, the chance of interference between the different memory traces will increase. Because older granule neurons are most likely already incorporated into other memory traces, activation of a granule neuron may result in retrieval of a memory trace that is irrelevant for the stimulus that the animal is exposed to. Since newly generated granule neurons are most likely not involved in multiple memory traces, incorporation of young neurons into a memory trace will reduce this type of interference. Computer models have also predicted that an increase in the number of stimuli a network is exposed to will lead to more mistakes in the output region, in the absence of neurogenesis (Deisseroth et al., 2004). However, continuous neurogenesis might also result in disturbed memory retrieval. Re-exposure to a familiar stimulus will reactivate the existing memory trace. Because of the low threshold for immature granule neurons to become activated, newly formed neurons that are not yet incorporated into a memory trace may also become stimulated and create a new memory trace. Besides the fact that this new memory trace is redundant, it might cause interference in the output region between the retrieved and the newly generated memory. In order to prevent this type of interference, the hippocampus may temporarily reduce the number of newly formed granule neurons. This is exactly what we observed when testing memory retention or reversal learning. Re-exposure to the Y-maze resulted in a significant reduction in the number of immature neurons in the DG (chapter 7). 126

128 General Discussion Potential mechanisms underlying the decrease in neurogenesis during memory retrieval The question remains which are the signals that induce a reduction in neurogenesis during retrieval of information. On the one hand, it can be an active process, which means that memory retrieval results in the upregulation of a specifi c signal in the DG which can actively suppress neurogenesis or promote apoptosis. On the other hand, memory retrieval might prevent certain factors from promoting neurogenesis and rescuing newly generated cells from apoptosis. Since the hippocampus is involved both in memory formation and retrieval, it has always been a topic of interest to understand how these two processes can be organized without interfering with each other. The most likely possibility is that there is a constant and tight restraint on synaptic plasticity in the DG, which can selectively be enhanced on detection of novelty (Grossberg, 1980). The DG can be considered as the gateway to the hippocampus. Only if the input to the hippocampus is suppressed under baseline conditions, continuous influx of information that may disturb memory retrieval can be prevented (Paulsen and Moser, 1998). This inhibition of granule neurons may be reduced upon detection of novel stimuli. Exposure to a novel environment, for instance, has been reported to cause a decrease in somatic inhibition of granule cells (Moser, 1996) and to enhance the induction and maintenance of LTP in the DG (Davis et al., 2004), which might support memory formation. The idea that synaptic plasticity in the DG is strongly suppressed under baseline conditions, and perhaps even more during memory retrieval, is supported by the fact that the DG is a brain region with a strong GABAergic activity. The mossy fi bers have many collaterals that innervate GABAergic neurons in the hilus, which project back to the granule neurons (Amaral, 1978; Buckmaster et al., 1996; Henze et al., 2000). The prominent inhibitory activity in the DG is illustrated by the observation that it is not possible to induce LTP in granule neurons without the addition of GABA-A receptor antagonists. However, granule neurons that are located near the proliferative zone, and therefore most likely represent newly generated neurons, do not require GABA-A inhibition for induction of LTP, indicating that newly formed neurons are less inhibited by GABA than mature granule neurons (Wang et al., 2000). Thus, newly generated granule neurons have a low threshold for becoming activated and their chance of fi ring is not under control of GABAergic input. These are features that increase the probability for interference between memory formation and retrieval. The only way to minimize the possibility of this type of interference is to reduce the total number of newly formed neurons during memory retrieval. As we have hypothesized in chapter 4, the survival of newly generated granule neurons may be negatively affected by lack of excitatory activity in surrounding neurons. Concomitantly, the prominent GAB- Aergic activity in the DG during memory retrieval may prevent newly generated neurons from long-term survival. Hence, we suggest that the hippocampus has two distinct states for memory formation and retrieval, depending on the level of GABA activity in the DG. The question 127

129 Chapter 8 remains which signals are responsible for the shift from one state to the other. It has been hypothesized that hippocampal levels of ACh may be a determining factor in this respect, with high levels during the encoding phase and low levels during retrieval (Hasselmo, 1999; Hasselmo and McGaughy, 2004). The concentration of extracellular hippocampal ACh rises when an animal is exposed to novelty, but remains low after repeated exposures to the same stimulus (Acquas et al., 1996; Orsetti et al., 1996). According to the results reported in chapter 4 and in the literature (Mohapel et al., 2005), ACh has no direct effect on survival of newly formed neurons. However, muscarinic receptor activation has been shown to cause a depolarization of granule neurons and to facilitate LTP in the DG (Burgard and Sarvey, 1990; Vogt and Regehr, 2001), indicating that ACh can modulate the balance between excitation and inhibition in the DG and therefore indirectly affect survival of newly generated neurons. However, it has also been suggested that the reduction in cholinergic input to the hippocampus during memory retrieval is regulated by the hippocampus itself, via hippocamposeptal connections from hippocampal inhibitory neurons to cholinergic and GAB- Aergic neurons in the MS. This would mean that the actual novelty detection takes place in the hippocampus and is not dependent on the input from the MS (Meeter et al., 2004). Incoming information is continuously compared with existing memory traces. Whenever there is a mismatch between incoming information and expectations based on the patterns that are present in the hippocampus, this might be recognized by either CA3 (Kesner et al., 2004; Lee et al., 2005) or CA1 (Lisman and Otmakhova, 2001) neurons and generate a novelty signal, which could have an inhibitory effect on DG granule neurons. Perhaps, the CA3 area is the most likely candidate to compare incoming information with existing memory traces. The EC projects to the DG, but also has direct connections with the CA3 area. Lesion studies indicate that memory retrieval remains undisturbed after disruption of the input from the DG to the CA3 area, but is impaired after damage to the perforant path axons that directly project from the entorhinal cortex to the CA3 region (Lee and Kesner, 2004). These fi ndings suggest that DG activation is avoided during memory retrieval. Potentially, feedback mechanisms from the CA3 area to the granule neurons, for instance via basket cells in the granule cell layer (Kneisler and Dingledine, 1995) or hilar mossy cells (Scharfman, 1994), further inhibit DG granule neurons if the stimulus is familiar or activate the DG in case the incoming information has not been encountered earlier. In summary, we here suggest that the reduction in hippocampal neurogenesis during memory retrieval might be mediated via a suppression of hippocampal activity during retrieval. Our data in chapter 4 suggest that an increased inhibition of the hippocampus reduces survival of newly generated neurons. However, the exact mechanism causing inhibition of the DG during memory retrieval still requires additional research, but it is not unlikely that novelty detection takes place in the hippocampal formation itself. 128

130 General Discussion Can this model explain supposed discrepancies in the literature? We can now use our model to look back at our own data and those in the literature to see whether we can explain the supposed discrepancies in the results that have been obtained. First of all, the question exists why some studies show changes in newly formed cell number upon acquisition of a hippocampus-dependent learning task and others do not. Gould and colleagues reported in 1999 that Morris water maze learning and trace eyeblink conditioning promote survival of newly generated neurons. Although we observed clear signs of increased morphological plasticity in newly formed neurons after Morris water maze learning, we could not reproduce the data obtained by Gould and colleagues, neither did we fi nd learning-induced increases in neurogenesis after other learning tasks. Also, in the literature there are various reports showing no effect of learning on the number of newly formed neurons (Snyder et al., 2005; Van Praag et al., 1999b). There is even evidence that spatial learning might decrease hippocampal neurogenesis (Ambrogini et al., 2004b) (see Table 1 in the General Introduction for an overview). Based on our model, we would not expect an increase in hippocampal neurogenesis after acquisition of a learning task in experimentally naive animals. The number of newly generated neurons available under standard conditions should be suffi cient for the acquisition of a single spatial learning task. It has been suggested that an increased survival of newly generated neurons upon training in a hippocampus-dependent learning task is an adaptation of the hippocampus to better cope with future challenges (Kempermann, 2002a). Although behavioral anticipation and also antcipation on the level of synaptic transmission are well-known phenomena, it is not very likely that the brain undergoes structural changes in order to be better prepared for a next stimulus. The brain primarily responds to the situation that it is in and it does not have the capability of predicting future situations. The survival of newly formed neurons may be enhanced when animals are exposed to multiple tasks. In that situation, the number of newly generated neurons may become a bottleneck and an increase in the capacity of the DG may be required in order to cope with the various stimuli that have to be encoded in different hippocampal memory traces. Perhaps, the rats that were used in the studies of Gould and colleagues were not experimentally naive. An increase in hippocampal neurogenesis in situations that require parallel processing of multiple stimuli will reduce the chance of interference between the different memory traces that are formed. Therefore, if animals were previously tested in other learning tasks, an increase in neurogenesis would have been expected during Morris water maze learning. Other studies reported a decrease in cell survival after Morris water maze learning. Dobrossy et al. (2003), for instance, showed that during the late, asymptotic phase of Morris water maze learning earlier formed cells die of apoptosis. Also Ambrogini and colleagues (2004) reported that Morris water maze learning in rats reduced hippocampal neurogenesis. The learning curve they presented shows that rats already acquired the task on the second training day and that training was still continued for another 3 days. If 129

131 Figure 7: Model on the role of newly formed granule neurons in memory formation. We propose that newly generated neurons facilitate memory formation, but that neurogenesis has to be reduced during memory retrieval in order to optimize the retrieval process and to prevent interference. Existing granule cells and newly formed neurons in the DG are presented by white and light grey circles, respectively. The target region of the DG, the CA3 area, and the hippocampal output regions CA1 and subiculum, are represented here as one group of cells, depicted as triangles. The dashed lines show existing, but not activated connections between granule neurons and CA3 pyramidal neurons. A memory trace is defi ned as the set of hippocampal neurons that become activated upon exposure to a certain stimulus. Highly activated neurons are dark red, dark blue or dark green; weakly activated neurons are pink or light blue. I) Baseline situation without exposure to a stimulus. II) Formation of memory trace A. Exposure to stimulus A, such as a learning task or a novel environment, will preferably activate newly formed granule neurons, since these are easier to excite and are receptive to plastic signals (Couillard-Despres et al., 2005; Schmidt-Hieber et al., 2004; Seki and Arai, 1999), but also some of the more mature granule neurons will be stimulated. Activation of a newly formed granule cell will lead to a stronger memory trace (shown in dark red) than activation of an older granule neuron (shown in pink). In the absence of neurogenesis, memory formation will still take place, but less effi ciently, resulting in a weaker memory trace. III) Formation of memory trace B. Subsequent exposure of the animal to stimulus B will, in this model, activate a diff erent set of cells in the DG and the output regions, although the memory traces partly overlap. The strong memory trace is depicted in dark blue; the weak trace is light blue. IV) Re-exposure to stimulus A will result in the most optimal retrieval (via memory trace A) if hippocampal neurogenesis is inhibited. Although there is some overlap with the retrieval of memory trace B, the output is most similar to the one of memory trace A. If the number of newly formed neurons is not reduced during retrieval, re-exposure to environment A will result in the simultaneous reactivation of the correct memory trace (memory trace A), but may also lead to the formation of a new memory trace (memory trace C, shown in green), which is superfl uous and can lead to interference between the retrieved and other memories (in this case memory trace B). Therefore, optimal retrieval is associated with the elimination of newly formed neurons that could cause this interference. Also, in the model without neurogenesis, problems may occur during retrieval. Granule cells that are activated upon exposure to environment A will activate memory trace A. However, since the granule cells that are activated upon exposure to stimulus A partly overlap with memory trace B, there is a chance that re-exposure to environment A will also activate memory trace B with the same probability, leading to possible interference between memory trace A and memory trace B and therefore retrieval of irrelevant information. 130

132 CA3/CA1 OUTPUT I NO STIMULUS II STIMULUS A: MEMORY TRACE A Neurogenesis No neurogenesis DENTATE GYRUS (DG) INPUT A to DG Strong memory trace INPUT A to DG Weak memory trace III STIMULUS B: MEMORY TRACE B Neurogenesis No neurogenesis = non-stimulated granule neuron = newly-formed granule neuron =non-stimulated pyramidal neuron = weakly activated neurons = strongly activated neurons = inactive connection = weak connection = strong connection INPUT B to DG Strong memory trace INPUT B to DG Weak memory trace IV RE-EXPOSURE TO STIMULUS A: RE-ACTIVATION OF MEMORY TRACE A Reduced neurogenesis No reduction in neurogenesis No neurogenesis INPUT A to DG Reactivation of the strong memory trace A INPUT A to DG Reactivation of memory trace A, but superfluous activation of memory trace C (in green) INPUT A to DG Reactivation of the weak memory trace A, but interference with memory trace B 131

133 Chapter 8 training is continued after asymptotic levels have been reached (overtraining), the memory will be stored in multiple sites of the brain (Markowitsch et al., 1985) and be less sensitive for hippocampal lesions (Akase et al., 1989). In other words, the memory has become hippocampus-independent. The same may be the case in the study of Dobrossy et al., where rats were trained for consecutive 8 days. This means that during daily re-exposure to the maze the formation of new memory traces would be redundant and could lead to interference with the earlier formed trace. The observed reduction in neurogenesis may therefore also serve to prevent interference between retrieved and newly formed memories. Interestingly, Dobrossy et al. reported a positive correlation between the rate of apoptotic cell death and reference memory. Animals that showed the highest levels of newly formed cell death performed best in the maze, supporting the idea that removal of new granule neurons facilitated memory retrieval. Studies in which neurogenesis was inhibited by irradiation or antimitotic drug treatment have shown that hippocampal neurogenesis is not a prerequisite for spatial learning. The only tasks that appear to be dependent on the presence of newly generated granule neurons are trace conditioning tasks (Shors et al., 2001; Shors et al., 2002). In these types of learning tasks, animals have to learn to connect two unrelated stimuli, the conditioned (CS) and the unconditioned stimulus (US), which are separated by a short interval, the trace. Perhaps, it is important for the acquisition of such a task that granule neurons in the DG that are activated upon exposure to the CS remain activated during the trace period. Considering the exceptional synaptic plasticity of newly generated granule neurons, this specific group of cells may possess characteristics that are appropriate, or even required, for the successful acquisition of this type of learning task Does neurogenesis have a function in humans? The experiments described in this thesis were all performed with rodents. Neurogenesis in rodents takes place at a relatively high rate. In a young adult rat, the daily production of new neurons is estimated to be around 9000 (Cameron and McKay, 2001). However, in animals with a more complex nervous system much less cells are generated in the hippocampus. In the macaque, for instance, only 200 new neurons are produced on a daily basis (Kornack and Rakic, 1999). Neurogenesis has also been detected in the human hippocampus (Eriksson et al., 1998), but due to the large spread in the age of death of the subjects that were studied and the variable time between BrdU injection and death, it is diffi cult to get an estimation of the total amount of neurons that is daily generated in the human brain. However, only 23% of the hippocampal BrdU-positive cells had a neuronal phenotype, suggesting that also in humans, hippocampal neurogenesis takes place at a low rate. This leads to the question whether the data on neurogenesis that were obtained in rodents are still relevant to humans. Perhaps, newly formed neurons have a distinct function in the adult rodent hippocampus, but lost its function in primates and humans. An important remark in this respect is that the primate and human brain samples that were analyzed for 132

134 General Discussion neurogenesis were derived from adult, or even aged, individuals. Neurogenesis is known to decline rapidly with age (Heine et al., 2004a; Kuhn et al., 1996; McDonald and Wojtowicz, 2005). The number of new granule neurons that is produced in a 24-month-old rat is only a fraction of what can be found in a 6-week-old animal. This age-related reduction in neurogenesis suggests that newly produced neurons may have a much more prominent role in hippocampal functioning in young animals than in middle-aged and aged individuals. Assuming that neurogenesis is involved in learning and memory, it appears logical that the production of new neurons is highest during the period that animals frequently encounter many new facts and situations. In primates and humans it may also be true that neurogenesis takes place at a high rate during the fi rst few months or years of life and that the newly generated neurons do have a distinct function during this period. Therefore, like in rodents, hippocampal neurogenesis in primates and humans may gradually become less important for hippocampal functioning during aging. 6. Recommendations for future research Although the data presented in this thesis provide some new insight into the potential role of newly formed granule neurons in learning and memory, many questions still remain to be answered. We showed that memory retrieval caused a strong reduction in the number of immature neurons. These results indicate that in future studies on learning-induced changes in neurogenesis, it may be interesting to shift the focus from memory formation towards other aspects of the learning process, such as consolidation, reconsolidation and retrieval. Moreover, it might be worthwhile to investigate the effects of different types of learning on hippocampal neurogenesis. So far, the Morris water maze has been used most frequently in neurogenesis research, but other learning tasks which are perhaps less stressful or in which the experimental animals experience a positive reward, instead of a negative stimulus, might also yield different points of view with respect to learning, memory and neurogenesis. Furthermore, the generation of new hippocampal granule cells during learning is most likely a very dynamic process. By choosing a single time point during or after the learning task at which levels of neurogenesis are evaluated, dynamic changes throughout the learning and consolidation process will not be detected. Termination of animals at different stages of the learning curve would provide information on potential differences in cell formation and survival during, for instance, the early phase of training compared to the late phase. It would be interesting to study whether the reduction in neurogenesis during retrieval occurs directly upon re-exposure to the familiar stimulus, or requires multiple retrieval sessions. In addition, the number of newly formed neurons might not always be the appropriate parameter to measure. If the number of new granule neurons is not altered by a certain learning task, it cannot be concluded that this population of neurons did not 133

135 Chapter 8 contribute to the memory formation. Perhaps, the involvement of newly formed neurons into a memory trace is not always reflected in changes in cell number, but causes changes in, for instance, the maturation of the dendritic tree, or the number of synapses that are made with neurons in the CA3 region. Morphometrical or electronmicroscopical analysis of newly formed neurons might be useful in this respect. In the model that we propose here, we presume that animals are still capable of learning in the absence of hippocampal neurogenesis, but that this results in a weaker memory trace. It is known that spatial learning in intact animals leads to increased immunoreactivity for certain kinases, such as PKCg in the CA1 area (Van der Zee et al., 1992) or for the PKA regulatory subunit 2 in the CA3 region (R. Havekes, personal communication). If animals without neurogenesis indeed form weaker memory traces, perhaps this can be visualized with immunocytochemical techniques, although the complex nature of a memory trace makes it diffi cult to determine the strength of a trace by staining of a single protein. Furthermore, we hypothesize that during retrieval of information there is a chance of interference, because there is always a certain overlap between memory traces. An increase in the number of memory traces, or in other words, an increase in the number of stimuli to which the animal is exposed, will increase the chance of interference. We therefore predict that animals without or with low levels of hippocampal neurogenesis will have diffi culties with multiple learning tasks within a short time window, and that animals with enhanced neurogenesis will perform better than animals with baseline levels of neurogenesis. Furthermore, we suggest that an increase in the number of tasks that the animal is trained in might promote the survival of newly formed neurons, in order to enhance the hippocampal capacity with highly plastic young neurons. We also found that wheel running strongly enhanced hippocampal neurogenesis and improved learning and memory. In order to test whether the facilitating effects of running wheel activity on different aspects of the Y-maze task are mediated by the increase in the number of newly formed neurons, an experiment could be performed in which the exercise-induced increase in neurogenesis is prevented. If the enhancement in neurogenesis is the underlying mechanism for the improved performance, exposing animals to cytostatic drugs or brain irradiation during the 14-day exercise procedure should prevent the benefi - cial effects of exercise on Y-maze learning. Possible complications with this experimental procedure are that antimitotic treatments also inhibit cell proliferation in other organs than the brain, which will make the animals feel sick and reduce their motivation for running or for fi nding food in the maze. A second restraint of reducing neurogenesis by these methods is that they also inhibit neurogenesis in the olfactory bulb, which may reduce olfactory perception (Lledo and Gheusi, 2003) and therefore interfere with the learning procedure. For the determination of the exact function of hippocampal neurogenesis in learning and memory, the use of transgenic animals in which neurogenesis can be stimulated or inhibited would be extremely useful. However, the problem is that many of the genes that are known to be expressed by any of the intermediate precursors during adult neurogenesis, 134

136 General Discussion such as nestin, vimentin or DCX, are also crucially involved in embryonic neurogenesis. Manipulating the expression of any of these genes will therefore result in malformations of the central nervous system or even in a lethal phenotype. The optimal solution would be to develop conditional knockouts for any of these genes. There are already existing mouse lines with disturbed neurogenesis that can be very valuable for the field, such as the GFAP-TK mouse (Garcia et al., 2004). As mentioned in the General Introduction of this thesis, neural stem cells are thought to be GFAP expressing radial astrocytes. In the GFAP-TK mice the herpes simplex virus thymidine kinase (HSV-TK) is expressed under control of the GFAP-promotor. Administration of the antiviral agent ganciclovir to the animal causes a selective ablation of the dividing GFAP-positive cell population. This treatment has been shown to reduce adult hippocampal neurogenesis by 98.2%, which makes the GFAP-TK mouse a very interesting tool to study the function of neurogenesis. Another mouseline that could be of potential interest is the ST8SiaIV-knockout mouse (R. Gerardy-Schahn, personal communication). ST8SiaII and ST8SiaIV are enzymes that are involved in the synthesis of polysialic acid (Hildebrandt et al., 1998). The former enzyme has a major role in NCAM polysialylation during embryonic development; the latter catalyzes polysialylation mainly in the adult brain (R. Gerardy-Schahn, personal communication). Genetic ablation of ST8SiaIV in a mouse might therefore provide an interesting model for studying the role of neurogenesis in learning, since it may be expected that ST8SiaIV-/- mice do not functionally incorporate newly generated neurons in the hippocampal circuitry. Since ST8SiaII-expression is not disturbed, embryonic development of the brain is unaltered in these animals. 135

137 Chapter 8 7. Concluding remarks Hippocampal neurogenesis is one of the most striking examples of plasticity in the adult central nervous system. It is a very dynamic process that is regulated by many factors, such as physical activity and the input from the MS. The addition of new neurons to the brain can increase the capacity of the DG, most likely without modifying existing synapses. Moreover, the exceptional capability of newly formed neurons for synaptic and morphological plasticity indicates that these cells may have a distinct role in optimal hippocampal functioning. Here, we hypothesize that newly generated granule neurons are positively involved in the formation of memory, but that a temporary reduction in the number of maturing neurons may improve memory retrieval. 136

138 General Discussion List of abbreviations ACh - Acetylcholine ASA - Active shock avoidance BrdU - 5-bromo-2 -deoxyuridine CORT - corticosterone CREB - camp-responsive element binding protein DCX - Doublecortin DG - Dentate gryus EC - Entorhinal cortex EDTA - Ethylenediaminetetraacetic acid EPSP - Excitatory postsynaptic potential GABA - Gamma-aminobutyric acid GAD - Glutamic acid decarboxylase GCL - Granule cell layer GFAP - Glial Fibrillary Acidic Protein LTP - Long-term potentiation MWM - Morris water maze MF - Mossy fi bers MPP - Medial perforant path MS - Medial septum NeuN - Neuronal nuclei NMDA - N-methyl-D-aspartate NSC - Neural stem cell RMS - Rostral migratory stream SGL - Subgranular layer SGZ - Subgranular zone SVZ - Subventricular zone TEC - Trace eyeblink condtioning ZT - Zeitgeber 137

139 Chapter 1, Figure 1 From neural stem cell to mature granule neuron Type 1 cell 2,5 / Radial astrocyte 1 D3 cell 1 D2 cell 1 D1 cell 1 GFAP+ 1,2,5 nestin+ 3,5 vimentin+ 4 small input resistance (< 500 MΩ) 5 self-renewing 1,2 PSA-NCAM+ 1 DCX+ 1 TUC-4+ 1 mitotic 1 PSA-NCAM+ 1,2 DCX+ 1,2 TUC-4+ 1 postmitotic 1,2 PSA-NCAM+ 1,2 DCX+ 1,2 TUC-4+ 1 postmitotic 1,2 putative stem cell 1,2,3,4 Type 2B cell 2 /Type Type 2A cell 2 II cell 5 /class 1 cell 6 Type 3 cell 2 Type 4 cell 2 NeuN+ 2 calretinin+ 2 nestin+ 2 nestin+ 2,3 nestin- 2 DCX+ 2 DCX- 2 DCX+ 2,3 DCX+ 2 postmitotic 2 mitotic 2 PSA-NCAM+ 2,5 PSA-NCAM+ 2 large input resistance (>500 MΩ) 5,6 mitotic 2 no spontaneous activity 5,6 mitotic 2,5 Mature granule cell NeuN+ calbindin+ postmitotic 138

140 Chapter 1, Figure 1 Figure 1: Summary of the diff erent subtypes of precursor cells that can be found in the hippocampal dentate gyrus during diff erentiation from a stem cell to a mature granule neuron. The neural stem cell (NSC) is thought to be a GFAP-positive cell with the morphological characteristics of radial glia. Throughout the diff erentiation process towards a mature granule neuron, diff erent subtypes of precursors can be defi ned. Seri and colleagues (2004) named these precursors D1, D2 and D3. D1 cells are mitotic; D2 and D3 cells are postmitotic. According to Kempermann and colleagues (2004) there is another subtype of cells, Type 2A cells, that is less mature than D1 cells (they are still DCX-negative), but that do not express GFAP anymore. Furthermore, Kempermann subdivided the DCX-positive mitotic cell population (D1 cells) into two subtypes: Type 2B and Type 3 cells. These two cell types are morphologically highly similar, but diff er with respect to their nestin expression. Cells that are positive for nestin and PSA-NCAM (Type 2B cells) were reported to have a very high input resistance (Fukuda et al. (2003); Ambrogini et al. (2004b)) and are also named Type II cells or Class I cells. During further maturation, cells reach the D2 stage, followed by the D3 stage. The diff erence between D2 and D3 is based on their morphology, with D2 cells having a short, thick process and D3 cells having a more elongated process. Kempermann and colleagues do not diff erentiate between these two cell types and gathered both of them under the same name: Type 4 cells. During the fi nal step of maturation, the cells express the Ca 2+ -binding protein calbindin, which is a characteristic of mature granule neurons. References: 1) Seri et al. (2004), 2) Kempermann et al. (2004), 3) Mignone et al. (2004), 4) Garcia et al. (2004), 5) Fukuda et al. (2003), 6) Ambrogini et al. (2004a). 139

141 Number of B rdu -positve i cells Ph e notype of BrdU-positive cells (%) Chapter 4, Figure 3 A ** *** B Hilus GCL C sham 30 nmol 60 nmol sham lesion D Hilus GCL E Nu m b e r o f k i-67 p o sitiv e c e lls NeuN GFAP unknown * F Hilus GCL 0 sham 30 nmol 60 nmol 140

142 Chapter 4, Figure 3 Figure 3: The plate illustrates the impact of the lesion on hippocampal neurogenesis. A) Infusion of either 30 nmol or 60 nmol of NMDA signifi cantly reduced the number of BrdU-positive cells in the GCL (*** P<0.001, ** P<0.01, post-hoc analysis following ANOVA). A representative image of BrdU-immunoreactive cells is shown in panel B. Scale bar=100 μm. The inset shows a higher magnifi cation of a few BrdU-positive cells in the inner blade. Scale bar=20 μm. C) The percentage of BrdU-positive cells that were double labeled for NeuN, GFAP or that did not display any of these two phenotypes did not diff er between lesioned (n=6) and sham (n=3) animals. Because there were no diff erences between the two lesioned groups, the data from the 30 nmol and the 60 nmol group were pooled. D) Example of a BrdU/NeuN double-labeled cell (arrow). The arrowhead points to a BrdU-positive cell that was neither NeuN nor GFAP-positive. BrdU is shown in green (DTAF), NeuN in red (Rhodamine Red) and GFAP in blue (Cy5). Scale bar=25 μm. E) The lesion did not cause changes in cell proliferation, as determined by Ki-67 immunocytochemistry. However, the two lesioned groups diff ered from each other. Infusion of 60 nmol of NMDA in the MS caused a signifi cant reduction of the number of proliferating cells compared to a lesion with 30 nmol of NMDA (* P<0.05, post-hoc analysis following ANOVA). F) The upper image shows Ki-67 positive cells in the dentate gyrus. Scale bar=100 μm. The lower image shows a magnifi cation of the selected area in the inner blade of the SGZ. Scale bar=10 μm. 141

143 Chapter 7, Figure 4 A B C D GCL GCL H H GCL GCL H H E GCL F GCL G GCL H I GCL H H H H J Figure 4: pcreb and DCX-immunoreactivity. DCX-expression is signifacntly increased after 14 days of exercise (B), compared to sedentary controls (A). C) Close-up of DCX-immunoreactive neurons in the inner blade of a mouse with enhanced neurogenesis after training in the Y-maze. D) Comparable picture as in C, but than for a mouse with enhanced neurogenesis after reversal learning. E) pcreb-immunoreactive cells in an animal with enhanced neurogenesis that was naive for the Y-maze. F) pcreb-positive cells in mouse with enhanced neurogenesis after memory retention. G, I) A strong colocalization was observed between pcreb (dark blue) and DCX (brown). H, J) Magnifi cation of the selected area in G and I, respectively. Black arrowheads point towards examples of DCX and pcreb-double positive cells, which can be distinguished from DCX-single labeled cells (white arrow) by their dark brown nuclear staining. Black arrows indicate pcreb-positive, but DCX-negative cells. H = hilus, GCL = granule cell layer. Scale bar: 100 μm for G and I; 50 μm for A, B, E, F, H and J; 25 μm for C and D. 142