| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Laboratoire de Physiopathologie des Comportements, Institut National de la Santé et de la Recherche Médicale, U588, Université de Bordeaux 2, Bordeaux, France
ABSTRACT I. NEUROGENESIS IN THE ADULT BRAIN: A NEW PARADIGM FOR STRUCTURE-FUNCTION RELATIONSHIPS II. FROM NEWLY BORN CELLS TO NETWORK A. Definitions B. In Vivo Evaluation of Adult Neurogenesis: Methodological Issues 1. Original studies: DNA synthesis staining and BrdU utilization 2. Phenotyping cells A) PRECURSOR CELLS. B) IMMATURE NEURONS. C) MATURE NEURONS. 3. Current identified pitfalls 4. Summary C. Integration of the Newly Born Cells Into Neurogenic Site Networks 1. The SVZ A) IDENTIFICATION OF THE STEMLIKE CELLS. B) MIGRATION OF THE NEWLY BORN CELLS THROUGH THE ROSTRAL MIGRATORY STREAM. C) LIFE AND DEATH OF THE NEWLY BORN CELLS. D) DIFFERENTIATION OF THE NEWLY BORN CELLS IN THE OLFACTORY BULB. E) FUNCTIONAL PROPERTIES OF THE NEWLY BORN CELLS. 2. The DG A) IDENTIFICATION OF THE STEMLIKE CELLS. B) LIFE AND DEATH OF THE NEWLY BORN CELLS. C) FATE OF THE NEWLY BORN CELLS. D) INTEGRATION OF THE NEWLY BORN CELLS. E) FUNCTIONAL PROPERTIES OF THE NEWLY BORN CELLS. 3. The cortex 4. Summary D. Factors Regulating Adult Neurogenesis 1. Intrinsic programs controlling neurogenesis A) CONTROL OF CELL PROLIFERATION. B) CONTROL OF CELL FATE. 2. Extrinsic factors regulating neurogenesis A) HORMONES AND NEUROSTEROIDS. B) NEUROTRANSMITTERS AND NEUROREGULATORS. C) TROPHIC FACTORS. D) MORPHOGENIC FACTORS. E) REGULATION BY GLIAL CELLS. F) REGULATION BY CELL DEATH. G) SUMMARY. III. THE ANATOMO-FUNCTIONAL APPROACH A. Environmental and Physiological Influences on Neurogenesis 1. Experience in enriched environments, neurogenesis, and learning A) OLD PROBLEM, NEW DATA. B) MECHANISMS INVOLVED IN ENRICHMENT-INDUCED NEUROGENESIS. 2. Reciprocal relation between learning and neurogenesis A) LEARNING INFLUENCES SEVERAL ASPECTS OF NEUROGENESIS. B) A BLOCKADE OF NEUROGENESIS DISTURBS LEARNING. 3. Discussion A) DO ADULT-BORN NEURONS PRODUCED BEFORE LEARNING HAVE A ROLE IN MEMORY PROCESSING? B) A ROLE OF CELL DEATH IN STABILIZATION? C) IS LEARNING-INDUCED CELL PROLIFERATION INVOLVED IN MEMORY CLEARANCE OR FLEXIBILITY? B. Activation of the Stress Axis: Acute and Long-Term Effects on Neurogenesis 1. Effects of stress on neurogenesis 2. Interindividual differences in stress reactivity 3. Perinatal manipulations C. Aging and Longitudinal Observations 1. Neurobiology of the age-related decline in neurogenesis A) INFLUENCE OF AGING ON NEUROGENESIS. B) FACTORS REGULATING NEUROGENESIS IN THE SENESCENT BRAIN. 2. Functional implications of age-related modulation of neurogenesis in memory D. Involvement of Neurogenesis in Pathologies 1. Pathological conditions associated with an upregulation of neurogenesis A) EPILEPSY. B) ISCHEMIA. C) HUNTINGTON'S DISEASE. D) OTHER TYPES OF CEREBRAL INJURY. 2. Pathologies associated with a downregulation of neurogenesis A) AFFECTIVE DISORDERS. B) SCHIZOPHRENIA AND NEUROLEPTICS. C) ADDICTION TO DRUGS OF ABUSE. IV. CONCLUSION ACKNOWLEDGMENTS REFERENCES
 |
ABSTRACT |
|---|
 Top NextReferences |
|---|
|
I. NEUROGENESIS IN THE ADULT BRAIN: A NEW PARADIGM FOR STRUCTURE-FUNCTION RELATIONSHIPS |
|---|
|
TopPreviousNextReferences |
|---|
The adult brain faces two seemingly contradictory challenges. On the one hand it must maintain behavior and thus preserve the underlying circuitry, and on the other hand, it must allow circuits to adapt to environmental challenges. This dilemma between stability and plasticity is a subject of debate. It has been suggested that the structural changes underlying plasticity exist at the levels of dendritic spines, dendrites, and axons (28, 287, 297, 383, 408). However, direct observation of dendrites and spines in the adult brain supports two divergent view points, i.e., both long-term dendrite stability (192, 368) and experience-dependent synaptic plasticity (541). Data suggest that dendritic spines are heterogeneous in shape and structure and can be roughly divided in two groups: large spines that survive for months and even years and small spines that are motile, change form rapidly, and either disappear or transform into large spines (for review, see Refs. 132, 216, 255). During long-term potentiation (LTP) of synaptic transmission, which is known to involve modifications in dendritic spine morphology (297, 337, 586), small spines are both generated and eliminated (143, 338). Given the structure-stability-function relationships, it has been suggested that the small spines are generated during activity-dependent processes and acquisition of memories, providing an inexhaustible source of new synapses, and that the large spines are the structural basis of memories in the long-term (255). For many authors, all these local and cellular phenomena have been labeled as neuroplasticity, a concept used to account for the functional observations.
In the search for structure-function relationships, great enthusiasm followed the discovery that embryonic neurons, for instance, dopaminergic neurons, could be transplanted into adult brains, survive and alleviate some postlesion behavioral alterations. Although it generated a considerable amount of data, this approach has not proven very successful so far in living up to functional and therapeutic hopes (157, 217). However, it has generated a considerable amount of data on the growth and the migration of these neurons, and on their ability to make synapse with the surrounding neurons of the host and to influence the activity of their target neurons. However, the turning point has been the demonstration that the adult mammalian brain is also capable of mitosis and that the generated cells differentiate into neurons that migrate to integrate circuitries. Interestingly, this particular new form of structural brain plasticity seems specific to discrete brain regions and most investigations concern the subventricular zone (SVZ) and the dentate gyrus (DG) of the hippocampal formation (HF), although an interesting debate developed in the wake of the proposition that neurogenesis occurs in the neocortex of the adult primate.
In brief, the discovery of neurogenesis in the adult brain confronted the persistent assumption that adult neurons did not undergo proliferation and that structure could not be changed in this way. In the 1960s, Altman's first observations of adult neurogenesis (11), followed by the studies of Kaplan and Hinds (for a historical view, see Ref. 249), were followed by negative reactions and critical publications that did not confirm the existence of newborn neurons in adults (452). Ironically, at about the same period, data were published on neurogenesis in birds related to the appearance of seasonal song, a discovery that gave rise to new thoughts on brain plasticity and adaptation (for review, see Ref, 410).
Overall, two main lines of research have emerged since the discovery that neurogenesis persists in discrete regions of the adult brain. The first aims to isolate neural stem cells and understand their fundamental biological properties, the ultimate objective being to manipulate them and enhance repair and regeneration. The second focuses on understanding its functional relevance, especially in the DG. We will thus approach these two lines of research in the following chapters. As we will see, it has now been unambiguously demonstrated that persistent neurogenesis occurs in the SVZ and DG of adult rodents. Most of our knowledge on the birth, proliferation, migration, and death of adult-born cells results from studies performed in the SVZ paradigm that easily allow the dissection of the properties of the stem cells (and their progeny) and the integration of the newly born neurons into preexisting networks. Once the existence of the phenomenon had been established, the important step was to understand what triggers and inhibits it, and how it is regulated. Identifying the functional roles of these adult-born neurons in the context of structure-function relationships will be the main focus of the review. The considerable evidence supporting the participation of these new neurons in brain functioning has been collected in the DG, and their role in information storage and hippocampal-related pathology will be examined. On the assumption that the DG transforms entorhinal inputs to facilitate storage in CA3, adult neurogenesis may allow the DG to adapt to new flows of relevant information while at the same time preserving it for the processing of old memories. In more general terms, we will ask whether neurogenesis acts on memory consolidation in a specific and functional manner or whether it prepares the HF for general experiences and challenges.
|
II. FROM NEWLY BORN CELLS TO NETWORK |
|---|
|
TopPreviousNextReferences |
|---|
The recent interest in adult neurogenesis studies has led to a promiscuous use of the term stem cell and a broadening of its definition. However, based on their functional properties, one can distinguish "stem" cells from "progenitor or precursor" cells. Thus a stem cell is currently defined as an undifferentiated cell that exhibits the ability to proliferate, to self-renew, and to differentiate into multiple yet distinct lineages (356, 570). In the adult brain, most stemlike cells are in a quiescent stage, except in the two neurogenic zones considered here, the SVZ and the DG of the HF, where they have a very slow dividing turnover (a few weeks). In contrast, progenitor cells are mitotic cells with a faster dividing cell cycle that retain the ability to proliferate and to give rise to terminally differentiated cells but are not capable of indefinite self-renewal; they are more committed than stemlike cells, and their multipotentiality is still a matter of debate. Finally, when the cell type being studied is not clear, as is the case in vivo, both stem and progenitor cells are referred to as precursor cells. For an overview of the characteristics that can be used to differentiate neural stem from progenitor cells, and of the controversies that still persist around definitions, we refer the reader to reviews by Gage (160), Geuna et al. (167), and Seaberg and van der Kooy (494).
B. In Vivo Evaluation of Adult Neurogenesis: Methodological Issues
1. Original studies: DNA synthesis staining and BrdU utilization
The original studies on in vivo cell proliferation relied on the use of [3H]thymidine ([3H]dT), which is incorporated into the cell during DNA synthesis, i.e., the S phase of the cell cycle. Mitotically active cells are subsequently revealed by autoradiography. More recently, BrdU, an analog of thymidine, has been used in lieu of [3H]dT (412). These markers have comparable availability times and labeling efficiencies (411). Although [3H]dT presents the advantage of good stoichiometry (allowing determination of the number of mitotic events after labeling), BrdU revealed by immunohistochemistry is more frequently used as the cost is lower, section processing is faster, and stereological analysis is possible. In addition, BrdU is easier to combine with other cellular markers, which makes it possible to phenotype the newly born cells by double or triple immunohistochemistry.
The original protocols used for BrdU labeling, established by Gage and colleagues (268, 267), consisted of a dozen BrdU injections (50 mg · kg–1 · day–1 ip). Since this pioneering work, with the flurry of data, protocols have diversified. As we will see in section III, the number and the doses of BrdU injections as well as the frequency of administration vary considerably from one experiment to another depending on the area of interest (fewer injections are required for studying the SVZ compared with the DG), the species, the phenomena being examined (a decrease or an increase in neurogenesis), and the subject type (young intact subjects, impaired or old individuals). To make the picture more complex, BrdU is sometimes administered in mice via drinking water (1–1.5 mg/ml for 2–4 wk) to minimize animal pain and distress (80, 332), with the disadvantage that individual daily intake is unknown.
Following BrdU injections, the killing of the animal is adjusted depending on the step of neurogenesis that is being investigated: cell proliferation, cell determination toward a given phenotype, cell migration and differentiation, or cell death. In this way, to study proliferation, animals are killed following the last BrdU injection within a short time lapse ranging from a few hours to a few days. As early as 2 h after a single injection, mitotic figures can be seen, and this time point has been proposed as a true measure of proliferation. However, in the case of senescent rats, multiple injections are required as cell proliferation and the frequency of labeled clusters are too low. Furthermore, when animals are killed within a short time after the last BrdU injection, newly born cells are small, irregular, and clustered (the clusters being more numerous with multiple BrdU injections), which renders quantitative analysis challenging. For this reason, a washout protocol of a few days allowing cells to migrate is applied in many experiments. Cell migration is followed by killing the animals at different days (1–10 days) after BrdU injection, in which case it is necessary to limit the number of BrdU injections (208). Survival of the newborn cells is usually studied by killing the animals after a long delay (several weeks) following BrdU administration, when the phenotype of the cells has been determined. Less frequently, retroviral tracing has also been used for "developmental" studies (81, 123, 439, 501). However, because retroviral infection is not controlled, labeling is quite variable and is thus unsuitable for quantitative analysis between different experimental groups (556). Finally, in most pharmacological and behavioral studies, the treatment under investigation is interrupted after BrdU pulses, and animals are killed after a "chase" period of several weeks. It is important to be aware that in this case, the influence of the treatment on cell determination and cell survival is not addressed. To do so, the "treatment" should be prolonged over the "chase" period time (for example, see Refs. 267, 335).
One of the most important tasks is to phenotype cells, and markers specific to the dividing cells or the state of maturation of newly born cells are available for in vivo studies.
A) PRECURSOR CELLS. So far, no marker that is exclusive to stemlike cells has been identified, which certainly reflects the divergence in opinion on the identity of the stemlike cell population. However, markers of stem-cell candidates or simply dividing cells are now available and can be divided into two categories depending on the purpose of the studies. The first category covers "universal" or general molecular markers that are used to define the biological characteristics of stemlike cells (for an extensive review, see Ref. 440). Among these markers, the most widely used is nestin, which defines classes of intermediate filament proteins; it was first described as the product of a gene whose expression distinguishes precursors from more differentiated cells in the neural tube (156) and was found to be specifically expressed in neural progenitor cells (307). More recently, a family of molecular markers has been identified, which includes mouse-Musashi1 (m-Msi1), a neural RNA-binding protein expressed predominantly in proliferating neuronal and/or glial precursor cells but not in newly generated postmitotic neurons and oligodendrocytes (246, 479).
The second category of markers is more specifically used in research aimed at uncovering the functional role of adult neurogenesis and consists of antigens present in cycling cells that are used as proliferative markers. They are often used in combination with BrdU, since incorporation of BrdU alone identifies cells undergoing DNA replication but does not formally provide evidence that these cells are actually capable of division. They include Ki-67, a nuclear protein expressed in dividing cells for the entire duration of their mitotic activity, the expression of which is neither linked to DNA repair nor to apoptotic processes (142, 259). Although less frequently used, proliferating cell nuclear antigen (PCNA) allows assessment of cell proliferation as well. It is an auxiliary protein of DNA polymerase 
which is increasingly expressed through G1, peaks at the G1/S interface, and decreases through G2 (146, 162, 294). This marker has been used successfully to study cell proliferation and cell cycle length in conjunction with BrdU cumulative labeling (39). Finally and much less used are P34cdc2, a key player in the initiation of mitosis (415), and phosphorylated Histone H3 (HH3) expression, which is confined to cells in the G2 and M phases of the cell cycle. In general, cells in the G2 phase exhibit punctuate HH3 nuclear staining that changes to a more condensed pattern once they enter the M phase (214).
B) IMMATURE NEURONS.
The mammalian RNA-binding protein Hu, the homolog of the Drosophila neuron-specific RNA-binding protein Elav, is exclusively expressed in postmitotic neurons (7, 479) and thus is frequently used as an early neuronal marker (31, 414). Other markers include neuron-specific class III 
-tubulin (TuJ1), a cytoskeletal protein expressed in all postmitotic neurons (360, 361), doublecortin (DCX), which encodes a microtubule-associated protein expressed in migrating neuroblasts (155, 168, 389), TOAD 64 (turned on after division, also called TUC-4, CRMP4, PUlip1), a cytoplasmic protein expressed transiently by postmitotic neurons (366), and the polysialylated-neuronal cell adhesion molecule PSA-NCAM (277, 498). Another candidate for immature neuronal markers is the basic helix-loop-helix (bHLH) transcription factor NeuroD, which seems to be expressed from a very early stage to the stage where new neurons develop dendrites (496, 497). Some of these markers stain nonneuronal cells, and some of them (TOAD 64, PSA-NCAM, DCX) are present in the brain, for example, in the piriform cortex, independently of neurogenesis (113, 389, 392, 423, 498, 560), which makes the interpretation of results more difficult.
C) MATURE NEURONS.
The most frequently used specific markers for mature neurons are neuron specific enolase (NSE), neuronal specific nuclear protein (NeuN) (386), and microtubule-associated protein (MAP-2). However, the identification of the neuronal phenotype of newly generated cells using these markers is considered ambiguous as they also label other types of cells (189, 411, 453, 454). An alternative approach consists in demonstrating that adult-born cells do not express any marker for glial cells such as 2',3'-cyclic nucleotide (CNP) for oligodendrocytes, and calcium binding proteins (S100, S100) or glial fibrillary acidic protein (GFAP) for astrocytes.
3. Current identified pitfalls
In addition to the problem of neuronal identification, several methodological pitfalls have been identified. First, embryonic and postnatal injections of BrdU have been reported to induce physical, morphological, and behavioral abnormalities (282, 495). Although only a few studies have addressed the deleterious effects of BrdU injections in adulthood, a neurotoxic effect has been reported in aged rats, loss of body weight, and deterioration in the state of the coat being induced by repeated administrations of 150 mg · kg–1 · day–1 of BrdU for 5 days (130). Together with its early teratogenic effects, this suggests that incorporation of BrdU into mitotically active cells may inhibit cell formation and affect stemlike cell population. This raises the as yet unresolved problem that BrdU may alter the functioning of the labeled cells. Second, changes in neurogenesis detected by variations in BrdU staining may be related to a modification in BrdU availability through an alteration in blood flow or in blood-brain barrier permeability. To rule out such confounding parameters, using endogenous markers of the cell cycle (Ki67, PCNA, HH3, and P34cdc2) and evaluating changes in different brain regions may be helpful. For behavioral studies, adequate control groups should be used to measure the involvement of a potential modification of blood flow by exercise. Third, pharmacological treatments or behavioral and environment-induced changes in neurogenesis may exert their effects by modifying the length of the cell cycle, the number of proliferating cells, or both. However, in most "functional" studies, evaluating the cell cycle parameters is difficult if not impossible, and one might question the relevance of this "exercise." Finally, BrdU or [3H]dT labeling might produce false negatives and positives. The most commonly used dose of BrdU (50 mg/kg) labels only a fraction of the proliferating cells (75), which may result in an apparent absence of effects of a given experimental setting on cell proliferation. However, multiple injections of this low dose may overcome this problem. Furthermore, because the precursor cell population most certainly does not divide synchronically, labeling all dividing cells with a single injection of BrdU, whatever the dose, appears utopist. On the other hand, higher doses of BrdU associated with an enhanced sensitivity of immunohistochemical methods might lead to labeling apoptotic cells or nonproliferating cells that synthesize DNA for repair (594), and thus to false positives. This problem is particularly relevant for studies on lesions, epilepsy, or ischemia, where a high rate of cell death is observed. There are several possible ways to rule out BrdU labeling of nonproliferative cells: 1) observing mitotic figures a few hours after BrdU or [3H]dT application; 2) using other endogenous cell cycle markers; 3) using electronic microscopy; 4) analyzing a time course of BrdU labeling (95); 5) combining BrdU with markers of cell death, although in this case false negative results due to the downregulation of markers by dying cells cannot be ruled out; 6) using retroviral labeling (81, 123, 439, 501, 556), which is not devoid of disadvantages (the type of cells incorporating the virus may not be fully controlled, and stereotaxic injection, known to cause injury, is required); and 7) visualizing neurogenic sites in nestin-promoter-GFP transgenic mice (577).
This last decade has been marked by tremendous advances including the development of BrdU labeling methods, the discovery of selective markers that differentiate neurons from glial cells, and the use of new molecular and genetic tools to follow the "development" of the newly born cells. As we will see in the following sections, this technical progress has led to a lot of research and data that have confirmed observations made in the early 1960s and considered at that time by many specialists as nonconclusive. However, researchers are now aware of problems and agree that proper interpretations depend on accepted rules: 1) [3H]dT and BrdU labeling are not sufficient to unambiguously differentiate old neurons from newborn cells, 2) the neuron-specific markers presently used must be complemented by appropriate controls to detect false positives and by electrophysiological recordings to assert real neuronal phenotypes, and 3) the maturation and the integration of the adult-born neurons must be followed up.
C. Integration of the Newly Born Cells Into Neurogenic Site Networks
A) IDENTIFICATION OF THE STEMLIKE CELLS. The SVZ, located throughout the lateral wall of the lateral ventricle, harbors the largest population of proliferating cells in the adult brain of rodents (12, 448, 511), monkeys (178, 179, 182, 248, 286, 434), and humans (47, 144), and it has been estimated that 30,000 cells are generated bilaterally daily in the mouse SVZ (317). This region exhibits a rostrocaudal gradient of proliferative activity: proliferation is higher in the dorsolateral corner of the rostral ventricle and falls caudally, moving from the striatum towards the HF (511). This gradient has been correlated with the capability of the constitutively proliferative cells to divide and to form neurospheres (381, 493) and certainly reflects the existence of two populations of dividing cells, one of quiescent stemlike cells and one of rapidly proliferating progenitor cells (381, 382). Four cell types have been described in the SVZ: 1) ependymal ciliated cells (type E) facing the lumen of the ventricle, whose function is to circulate the cerebrospinal fluid; 2) proliferating type A neuroblasts, expressing PSA-NCAM, Tuj1, and Hu, and migrating in "chains" toward the olfactory bulb (OB); 3) slowly proliferating type B cells expressing nestin and GFAP, and unsheathing migrating type A neuroblasts; and 4) actively proliferating type C cells or "transit amplifying progenitors" expressing nestin, and forming clusters interspaced among chains throughout the SVZ (15, 55, 122, 124, 166, 220, 474).
Although it has been proposed that the stemlike cells may be ependymal type E cells, which were shown to divide in vivo and to differentiate into neurons in the OB (242), this view has been challenged by van der Kooy and colleagues (85) who have found that ependymal cells generating spheres do not have the ability to self-renew or to produce neurons, and by Capela and Temple (79) who have shown that ependymal cells do not form neurospheres. It is more likely that the type B cells represent a stemlike cell type (123, 124). Indeed, after 1 wk of intracerebroventricular (icv) infusion of the antimitotic drug cytosine--D-arabinofuranoside (AraC), type C and type A cells are eliminated, whereas type B cells continue to divide. Between 1 and 2 days after treatment cessation, type C cells reappear, followed 2 days later by type A cells, suggesting that type B cells give rise to type C cells, which in turn generate neuroblasts (125). Furthermore, following the targeted introduction of a retrovirus into GFAP-positive cells, which include type B cells, labeled cells migrate towards the OB where they give rise to new neurons (123). Thus the lineage progression B 
C A has been proposed (123, 124). However, Doetsch et al. (126) have also shown that after exposure to high concentrations of EGF, the type C cells retain stemlike cell properties (126). In fact, there are opposing views on whether neurospheres are derived from GFAP-positive cells (type B cells) or from fast proliferating type C cells (225, 380) (see Fig. 1).
|
After reaching the middle of the OB, the newborn cells detach from chains, migrate radially, and progress into one of the overlying cell layers whereupon they undergo terminal differentiation. Neuroblast detachment from chains is initiated by Reelin and tenascin, whereas radial migration depends on tenascin-R (198, 478).
C) LIFE AND DEATH OF THE NEWLY BORN CELLS.
Massive cell death has been observed during the first 2 mo after a BrdU pulse (572). This elimination mechanism is prominent in the OB compared with the RMS and the SVZ (39, 50, 439) and may maintain constant OB cell number by a continuous cell turnover, as was suggested during earlier development (419). The number of newly born cells that survive 1 mo after a BrdU pulse (50 mg · kg–1 · day–1 of BrdU for 4 consecutive days in 2-mo-old female wistar rats) has been estimated to be 60,000 per granule cell layer in one study (50) and 120,000 in another (572). In the latter case, it was shown that 50% of the newly generated neurons (
80,000 per granule cell layer and 800 per periglomerular layer) that survived the initial period of cell death survived for at least 19 mo (572), confirming earlier work (253). With the use of retroviral labeling of precursors in the SVZ, it was confirmed that one-half of the labeled cells died shortly after their arrival in the OB (between 15 and 45 days after neuronal birth) and that most dying cells were mature, harboring dendritic arborization, and receiving connections (439). It was further shown in this study that survival of the newly generated granule cells depends on sensory input.
D) DIFFERENTIATION OF THE NEWLY BORN CELLS IN THE OLFACTORY BULB. The newly born cells differentiate mainly into neurons that grow dendritic trees and differentiate into two types of intrabulbar interneurons. Most of the cells (75–99%) differentiate into GABA granule cells (GCs), whereas a smaller number (1–25%) differentiate into periglomerular cells expressing GABA and/or tyrosine hydroxylase (38, 81, 256, 326, 439, 475, 572). Several months after their birth, 10,000 GABAergic granule neurons, 100 periglomerular GABAergic neurons, and >60 new dopaminergic periglomerular neurons have survived and are added per bulb daily (572).
The different maturation stages of adult-born granule cells have been observed using retroviral labeling of precursors in the SVZ (439). Five different classes of adult-born cells were distinguished according to their morphology and location: 1) migrating neuroblasts in the rostral extension of the RMS (days 2–7), 2) neuroblasts migrating radially (days 5–7), 3) GCs with dendritic processes that do not extend beyond the mitral cell layer (days 9–13), 4) GCs with a nonspiny dendritic arborization in the external plexiform layer (days 11–22), and 5) mature GCs with extensive dendritic arborization (days 15–30).
The newly born GCs and periglomerular neurons appear to be synaptically integrated into the existing circuitry as they are labeled following an injection (within the piriform cortex) of a green fluorescent protein (GFP) expressing pseudorabies virus (PVR GS518) known to be transported along the neurons and to cross synapses (80).
E) FUNCTIONAL PROPERTIES OF THE NEWLY BORN CELLS. The temporal sequence of electrophysiological changes in the adult born neurons has been observed by coupling live imaging (the newly born cells being labeled with a replication-defective retrovirus expressing eGFP) and patch-clamp recordings (81). Migrating neuroblasts [class 1 and 2 according to the description of Petreanu and Alvarez-Buylla (439)] were found to be silent, action potentials appearing later. Class 1 cells expressed functional GABAA receptors and AMPA receptors, and when they started migrating radially (class 2), they expressed functional N-methyl-D-aspartate (NMDA) receptors. Synaptic (GABAergic then glutamatergic) inputs were present in nonspiking neurons (class 3 and 4) while they were growing their dendrites. Spiking activity was the last property acquired by class 5 neurons, their electrophysiological characteristics being indistinguishable from those of older GCs.
The functionality of adult-born neurons was further examined using the induction of the immediate early gene product c-Fos, a marker for mapping activated neurons (222). With this approach, it has been shown that 1) in mice, 3- to 7-wk-old adult-born periglomerular neurons respond to physiological (odors) stimuli (80); and 2) in hamsters, sociosexual cues are able to activate the adult-born neurons localized in both the accessory and the main OB (221).
A) IDENTIFICATION OF THE STEMLIKE CELLS.
Cell proliferation was first demonstrated in the DG of rodents 40 years ago by autoradiography in a germinal zone which is not, in contrast to the SVZ, located close to the walls of the ventricle. This subgranular zone (SGZ) is located at the interface between the granule cell layer (GCL) and the hilus of the DG, deep within the parenchyma (12, 13). This cell proliferation is also a feature of monkeys (178, 182, 284) and humans (144). The nature of the proliferating cells is still a matter of debate. According to work carried out in Alvarez-Buylla's laboratory (501), the stemlike cells may correspond to a subpopulation of GFAP-positive cells, equivalent to the type B cells described in the SVZ, since 2 h after their birth a large majority of newly born cells express GFAP (501). While the number of these cells decreases in the following days, the proportion of dividing GFAP-negative cells increases. These small dark cells, originally described by Kaplan and Bell (251), are called D cells. Chronic treatment with the antimitotic AraC eliminates most dividing cells, D cells, and astrocytes, but soon after treatment cessation, some surviving B cells begin to divide, whereas D cells reappear only at 4 days. Type B cells, specifically labeled with an avian retrovirus, give rise to granule neurons with mossy fibers reaching the CA3 subfield. These results indicate that B cells act as stemlike cells, regenerate D cells, which function as transient precursors, and give rise to new granule neurons. This hypothesis is reinforced by the observation that astrocytes retain expression of m-Msi1 (480). On the basis of the analysis of nestin-promoter GFP transgenic mice, two classes of cells have been identified: type 1 cells (GFAP+, S100–, DCX–, PSA-NCAM–), which correspond to type B cells, and are the putative stem cell, and type 2 cells (GFAP–, S100–, DCX+, PSA-NCAM+) supposed to be the type D cells (154, 159, 289, 521). Because there is no overlap between glial and neural markers in type 2 cells, an observation inconsistent with Seri et al. lineage (501), it has been proposed that either type 1 cells divide asymmetrically and thus generate a neuronal lineage-restricted progenitor cell (type 2) and a glial lineage-restricted progenitor cell, or alternatively that a transient stage (lacking glial and neuronal markers) exists between the type 1 and type 2 cells (521). On the other hand, Palmer et al. (424) reported that very few proliferating cells in the adult DG are GFAP-positive (424), a discrepancy as yet unexplained. Finally, the assumption that the DG harbors stemlike cells has been challenged by showing that this region contains progenitor cells with restricted properties rather than stemlike cells with self-renewing properties (475, 493) (see Fig. 2).
|
24 h in 3-mo-old rats (75). It has been proposed that the progeny continue to divide further during the first week after their birth as the number of labeled cells continues to rise (204). Recently, it has been confirmed that the cells are still in the cell cycle 3 days after their initial division (521). The clusters of dividing cells have been shown to include several cell types: neural precursors, committed neuroblasts, glial precursors, and endothelial precursors (angioblasts representing 37% of proliferative cells) that are proximal to vessels (256, 363, 424). This suggests that neurogenesis most probably occurs within the context of vascular recruitment. Within this niche, endothelial cells may constitute a critical regulator for self-renewal of stemlike cells (505) through the release of trophic factors (see sect. IID2C and Table 1).
|
9,000 newborn cells were found to be generated per day (75). In a more recent study, only 4,000 new cells, out of which 3,000 were found to be new neurons, were shown to be added daily in the DG of 4-mo-old rats (456); this discrepancy in the number of newborn cells is certainly linked to the difference in the age of the animals used in the studies, as the production of new neurons in the DG diminishes with age (see sect. IIIC1A). In the macaque, the number of newborn cells per day (200) represents 0.004% of one GCL (284).
C) FATE OF THE NEWLY BORN CELLS.
The newborn cells differentiate mainly into neurons in the GCL. In his original study using electron microscopy, Kaplan and co-workers (249–252) reported that the newborn cells exhibit the ultrastructural characteristics of neurons. Then, they were shown to express neuronal markers. During the first 2 days after their birth, most BrdU-labeled cells express nestin (95), then TuJ1 (75) or TOAD-64 (75, 536) in the following 3 days, and DCX between 1 and 14 days after their generation (65). DCX-BrdU-colabeled cells show features of progenitor cells as some of them coexpress Ki-67 and are thus still able to divide (58, 154, 521). This rapid postmitotic status of new cells has been confirmed with calretinin, a calcium binding protein that is transiently expressed by postmitotic cells, the expression of which is observed as early as 1 day after labeling dividing cells, a peak being reached after 1 wk (58). Although newly born cells express NeuN as early as 72 h after their generation (58, 193), only half of the BrdU-IR cells expressed NeuN over 3 wk (65). Using NSE, half of the newly born cells were found to be neurons 2 wk after their birth (78). Calbindin is expressed later, by 3 wk after birth (180, 290, 536, 537). A low percentage of adult-born cells differentiates into astrocytes (GFAP/S100), and rare are those that adopt a microglial or oligodendrocyte phenotype (521). It should be emphasized that even 4 mo after their labeling a nonnegligible proportion of BrdU-IR cells have an unknown phenotype. In monkeys, most newly generated cells are claimed to be neurons as well since they express TuJ1, TOAD-64, NeuN, NSE, and calbindin, and rarely markers of astrocytes (GFAP) or oligodendrocytes (CNP) (178, 182, 284). It was first thought that the newly born cells did not survive, thereby constituting a continuously refreshed pool of neurons (189). However, it was more recently found that they do not have an ephemeral existence and survive for at least 11 mo (263).
Recently, it has been shown that beside glutamatergic granule neurons, a small percentage of newly born cells (14%) differentiate into GABAergic basket cells (316). Furthermore, under specific circumstances (ischemia or neonatal kainate administration), pyramidal cells in Ammon's horn are also generated (see also sect. IIID1B; Refs. 128, 399, 489). However, the progenitors responsible for the regenerating pyramidal neurons originate from the periventricular region rather than from the SGZ (399).
D) INTEGRATION OF THE NEWLY BORN CELLS. The newborn cells integrate the GCL 4–10 days after their generation. As they form dendrites, they receive synaptic contacts and extend axons into the CA3 region, suggesting that they make synapses long before being fully mature (75, 204, 252, 344, 517). Indeed, using virus-based transynaptic activity neuronal tracing (PVR GS518), the neurons generated in the DG have been shown to be synaptically integrated into the preexisting circuitry 4–8 wk after their birth (80), whereas they reach a mature morphology (soma size, total dendritic length, dendritic branching, and spine density) only 4 mo after their birth (513).
E) FUNCTIONAL PROPERTIES OF THE NEWLY BORN CELLS. When still located at the interface of the hilus, the newly born cells exhibit electrophysiological properties characteristic of immature neurons: they are completely unaffected by GABAA receptor inhibition, and they exhibit paired-pulse facilitation, have a lower threshold for induction of long-term potentiation (LTP), and display robust LTP (564). With the use of nestin-promoter GFP transgenic mice, type I cells (putative B cells) were found to have low input resistance values, whereas the type II cells (type D cells, expressing PSA-NCAM) exhibited higher input resistance and voltage-dependent sodium currents (159). Recently, PSA-NCAM expressing cells have been shown to differ from mature neurons in their passive (input resistance) and active membrane properties (such as calcium spikes that boost fast sodium action potentials) and in their enhanced ability to develop LTP (490). One month after their birth, the "newborn" neurons, labeled by a retroviral vector expressing GFP, exhibit some electrophysiological properties (input resistance, threshold potential for spiking and firing) similar to those of mature granule neurons (556). The stimulation of the perforant pathway, the main excitatory afference to the DG, elicits responses in "newborn" GFP-labeled cells, indicating that they receive functional synaptic inputs and are therefore functionally integrated into the preexisting network. However, a depression of synaptic currents is evoked in these cells following the paired-pulse stimulation of the perforant pathway, whereas a facilitation response is observed in mature cells, a discrepancy attributed to differences in presynaptic release properties (556). The functionality of the adult-born granule cells has been further demonstrated by showing, using c-fos, that 40% of 1-mo-old newborn granule neurons were able to respond to various chemoconvulsants (232, 234). In contrast, following a more physiological form of hippocampal stimulation consisting in a hippocampal-dependent learning task, only 3% of cells are activated (232). Finally, besides these functional excitatory adult-born granule cells, it has been shown that GABAergic newborn basket cells form functional inhibitory synapses with the granule cells (316). This discovery raises the question of the mechanisms and factors involved in the production of excitatory granule neurons versus inhibitory interneurons.
Although very controversial, the existence of adult neurogenesis in the cortex deserves attention. Indeed, Altman and co-workers in the 1960s (11, 12, 14) and Kaplan in the early 1980s (247–249) reported evidence for neurogenesis in the cortex of rats and cats. More recently, neurogenesis has been described in the rat anterior neocortex (182) and in the prefrontal, inferior temporal, and posterior parietal cortex of macaques (179, 182). Within 2 wk after BrdU injection(s), the number of BrdU-labeled cells increased and then fell after 5 wk, indicating that a large number of newly born cells died. The surviving cells were assumed to be neurons since they extended axons locally and expressed markers of immature (TOAD 64, TuJ1) and mature (38–52% of BrdU-NeuN cells) neurons but did not express GFAP or CNP. Although the site of origin of the newly born neurons remains unclear, it has been proposed that they may be generated into the SVZ and migrate to neocortical areas through the white matter or may be recruited from local quiescent stemlike cells (179, 182, 284).
In contrast to these data, other groups have reported a complete absence of neurogenesis in the mouse (134, 332) and monkey cortex (281, 453). Neurogenesis was observed in the neocortex of mice only following a targeted apoptotic degeneration of corticothalamic neurons (332), and in this case, endogenous neural precursors migrated and differentiated into neocortical neurons in a layer- and region-specific manner and reformed appropriate long-distance corticothalamic connections; thus they appeared to reconstruct the lesioned circuits. In the primate neocortex, although cell proliferation occurs, the newly born cells do not seem to differentiate into neurons (285). These discrepancies may be due to differences in housing conditions, the animals' histories, and genetic background, and other technical considerations (survival times, BrdU dosage and administration mode, immunohistochemistry...), but generally, the existence of this cortical neurogenesis is still a matter of debate (411).
Following a century of doubt and controversies, there is now a consensus that neurogenesis occurs in the adult brain in at least two regions, the SVZ and the DG. In both structures, stemlike cells proliferate, migrate, and differentiate mainly into granule neurons that will synthesize GABA in the OB and glutamate in the DG. Although less studied, a small proportion of adult-born cells differentiate into other types of interneurons (the periglomerular in the OB and the basket cells in the DG). Altogether this adult neurogenesis leads to the birth of 30,000 new neurons per day in the SVZ and between 3,000 and 9,000 in the DG of young adult rats, depending on their age. The reasons for which, 1) the SVZ and the DG harbor adult neurogenesis, 2) neurogenesis is curtailed in the DG compared with the OB, and 3) genesis rates are much lower in primates, are currently unknown. Furthermore, because these structures do not grow in size, a homeostatic compensatory equilibrium must be attained through an increase in cell death that must be equivalent to the initial addition of neurons. This poorly understood phenomenon, in particular in the DG, deserves more attention for it is an important partner of neurogenesis. Finally, recent evidence indicates that the adult-born neurons of the OB and the DG are functional and thus play a physiological role. Although these findings suggest a relevant contribution of these newly generated neurons to the bulbar or hippocampal function, further studies are needed to confirm these reports and fully unravel their fundamental consequences on the animals' behavior (see sect. III).
D. Factors Regulating Adult Neurogenesis
Although various factors that affect the division, migration, and differentiation of neural precursor cells have been isolated, the precise mechanisms that control neuronal fate in the adult nervous system remain largely unknown. Both cell intrinsic programs and extracellular/environmental factors, which we will review below, are at play.
1. Intrinsic programs controlling neurogenesis
Two fundamental decisions are involved in order for a precursor cell to generate a neural cell. The first one is to decide whether to self-renew or to undergo mitotic arrest, and the second is to interpret mitotic arrest, using cell autonomous cues that direct toward a particular fate. Much of our current understanding of these cell intrinsic programs comes from work performed when neurogenesis is at its best, i.e., during development. However, it is important to emphasize that the rules that govern neuronal specification during development may not be the same in the adult brain. Furthermore, since a description of the detailed molecular mechanisms involved in cell proliferation and determination is beyond the scope of this review, we here propose only a brief overview of the intrinsic factors involved, with a special mention whenever their involvement has been extended to the adult neurogenic zones.
A) CONTROL OF CELL PROLIFERATION. Among the cell cycle factors regulating cellular proliferation, Rb (retinoblastoma) and its related proteins (p107, p130), necdin, and the E2F protein families are key players (for a review, see Ref. 580). Thus, during the G1 phase, Rb predominates in a hypophosphorylated form that can bind to E2F, a positive regulator of the cell cycle (403), thereby repressing its transcriptional activity and preventing the cells from entering the S phase. In cycling cells, phosphorylated Rb accumulated during the late G1 phase releases E2F, thus allowing S phase entry. This phosphorylation depends on the activation of cyclin-dependent kinases (CDK) acting sequentially. Early to mid-G1, cyclins of the D class (D1, D2, and D3) activate CDK4/CDK6, and in late G1, cyclin E activates CDK2, leading to hyperphosphorylation of the Rb protein. Two families of CDK inhibitors (CDKIs) that can suppress cell proliferation by inhibiting Rb phosphorylation have been identified (566): members of the inhibitors of CDK4 family (INK4 family), including p15Ink4b, p16Ink4a, p18Ink4c, and p19Ink4d, and members of the kinase inhibitory protein family (Cip/Kip family) such as p21Waf1/Cip1, p27Kip1 or p57Kip2 (see Fig. 3).
|
The role of cell cycle regulators in the control of neuron production has been studied as well. Transgenic mice that lack p27Kip1 expression display a higher rate of cell proliferation versus differentiation in the SVZ, leading to an increased number of type C cells, a reduced number of type A neuroblasts, and no change in the number of type B cells (127). Finally, distinct functions for CDK inhibitors, either in the control of cell cycle exit and differentiation during neurogenesis (respectively, p27Kip1 and p19Ink4d) or in the maintenance of a quiescent state in neural progenitors (p18Ink4c) or neurons (p21Cip1) in adults have been underlined (303).
B) CONTROL OF CELL FATE. Data from developmental biology have clearly indicated that a core genetic program involving multiple bHLH transcription factors is required for both neuronal differentiation and determination. These factors, deriving from proneural and neurogenic genes, antagonistically control the switch from cell proliferation to neural differentiation: cascades of neuronal bHLH genes promote differentiation, whereas antineuronal bHLH genes repress them under the control of Notch and keep cells at a precursor stage (for recent general reviews, see Refs. 358, 473).
Two classes of proneural genes can be distinguished: the determination factors, such as Mash1 (mammalian achaete-scute homolog), Math1 (mammalian atonal homolog), and Ngns (Neurogenins) including Ngn2, expressed early in mitotic neural precursor cells, and the differentiation factors, including NeuroD, NeuroD2, and Math2, expressed later in postmitotic cells (for reviews, see Refs. 49, 379). It was recently determined that these proneural genes are downstream effectors of Pax6 (for a review, see Ref. 174), a transcription factor that promotes neurogenesis (210).
These proneural genes are components of a cell-cell signaling mechanism whereby a cell that becomes committed to a neural fate inhibits its neighbor from doing likewise. This process of lateral inhibition, which restricts the domains of the proneural gene activity and involves neurogenic genes, is mediated by the Notch pathway (for reviews, see Refs. 23, 32, 161, 244, 491, 567). Among the effectors of Notch, three families of negative regulators of the bHLH transcription factors are known: the HES, Id, and HES-related (HESR or HERP) families (226, 413, 481, 545). These effectors repress neuronal determination and differentiation in those cells not destined to become neuroblasts. Paralleling this regulation of neuronal proliferation by inhibiting neuronal fate, Notch was also found to be involved in determining glial fate (186, 398, 467).
Recent studies have highlighted that some of these factors, which are at play during development, may be involved in neuronal cell fate during adulthood. Thus Notch1 and Hes5 were found to be expressed at high levels and with a similar pattern in both the adult SVZ and DG, whereas numb and numblike, which negatively regulate the Notch signal transduction, were not, suggesting an active Notch signal in the adult neurogenic zones (523). Similarly, the mRNA expression of several bHLH factors was found to be present at various degrees in the adult HF, ranging from the restricted expression of Mash1 within the proliferative SGZ, which is consistent with its role in maintaining a precursor cell phenotype, to the widespread profile of Hes5 throughout the HF (140).
In conclusion, despite the considerable advances achieved recently with the development of molecular tools, the complex intrinsic machinery that controls and coordinates proliferation and differentiation is still poorly understood. As is the case for hematopoiesis, for which much progress has been achieved, understanding the gene expression patterns of progenitor cells and their progeny will be a critical step in elucidating the mechanisms underlying cell fate.
2. Extrinsic factors regulating neurogenesis
A) HORMONES AND NEUROSTEROIDS. I) Adrenal corticosteroids. Historically, corticosteroids were the first factors to be studied for their influence on adult neurogenesis (354). Corticosteroids are released into the blood circulation following the activation of the hypothalamo-pituitary-adrenal (HPA) axis, primarily by stress (351). Corticosterone, the main corticosteroid in rodents, regulates its own secretion through negative feedback, by interacting with two receptors (the mineralocorticoid MR, the glucocorticoid GR), present in the DG. Suppression of corticosterone secretion after bilateral adrenalectomy (adx) increases glial and neuronal births in the DG (71, 176), whereas mitotic activity in the SVZ remains unchanged (465), suggesting a site-specific inhibitory influence of corticosteroids. It was further shown that proliferation increases within 24 h after adx and remains constant over the 6 subsequent days; the newly generated cells survive for at least 4 wk in the absence of corticosterone, indicating that their survival is corticosterone independent (465). Cell death is also enhanced, but the populations of cells undergoing mitosis or apoptosis are distinct: immature cells divide at the interface of the hilus, whereas more mature neurons, located at the interface of the molecular layer, die (72). These alterations are prevented by corticosterone replacement (176, 465). The respective roles of MR and GR on adrenalectomy-induced structural modifications have recently been examined (374). Treatment with a low dose of the MR agonist, aldosterone, prevents adrenalectomy-induced increase in cell death, whereas a higher dose is necessary to normalize cell proliferation. Furthermore, treatment with a GR agonist, RU 28362, at doses that should fully occupy this receptor prevents both adrenalectomy-induced cell death and birth. Thus the stimulation of both MR and GR mediates the effects of corticosterone on cell proliferation and protects mature cells from cell death.
This inhibitory action of corticosterone on neurogenesis has also been found in other models: 1) acute corticosterone administration {2 x 40 mg/kg administered 1 h before a single [H3]dT injection (71, 76); implantation of a 200-mg corticosterone pellet followed by 1 x 200 mg/kg BrdU (17)}, 2) chronic corticosterone treatment [1 x 40 mg/kg for 21 days; BrdU 10 x 50 mg/kg every 12 h on days 17–21 (212)], 3) stress (see sect. IIIB1), and 4) interindividual differences in the activity of the HPA axis (305; see sect. IIIB2).
The mechanisms by which corticosterone hampers cell proliferation remain unknown. In vitro, glucocorticoids block the G1 phase of the cell cycle, notably through repression of cyclin D1, cyclin D2, Cdk4 and Cdk6, and induction of the CDKIs p21Cip1 and p27Kip1 (438, 585, 592). However, it remains to be determined which of these mechanisms is involved in vivo. Because only a small minority of precursor cells express corticosteroid receptors (77), corticosterone may not act directly on proliferating precursors, although it cannot be excluded that immature forms of receptors that are not recognized by current antibodies are involved. Corticosterone could act on neighboring glial or neuronal cells expressing corticosteroid receptors, which could control the cell cycle by releasing growth factors (515). Alternatively, corticosterone could regulate proliferation indirectly by increasing glutamate release in the DG (520) (see sect. IID2B) as the effects of adx or high levels of corticosterone can be blocked by NMDA receptor activation or inactivation, respectively (76). Corticosterone may also inhibit cell proliferation by dowregulating the production of insulin-like growth factor I (IGF-I) (585, 592).
II) Gonadal hormones.
The influence of female gonadal hormones has been investigated in the DG since estrogen replacement therapy seems to reduce the risk of age-related cognitive impairments (213). Although cell proliferation in the GCL (and not the hilus) is higher in female than in male rats, the newly born cells do not survive, which explains the lack of sex differences in the number of BrdU-IR cells 2 wk after labeling (536). Sex-dependent proliferative activity involves the stimulatory influence of estrogens, since the number of BrdU-labeled cells is highest during the proestrus phase of the estrus cycle, when circulating levels of estrogens are highest (536), and acute administration of 17-estradiol reverses the ovariectomy-induced decrease in cell birth (29, 536). In contrast, using TOAD-64 and calbindin as early and late neuronal markers, it has been shown that estradiol does not alter neuronal differentiation (536).
However, discrepant results have been reported following chronic administration of estradiol to ovariectomized adult female rats (435), spontaneous hypertensive rats (436), or adult wild meadow voles (163). Cell proliferation in the GCL (measured 24 h after a single intraperitoneal injection of [3H]dT) is lower in female meadow voles captured during the breeding season, when estradiol levels are high, compared with reproductively inactive females. A similar relationship has been confirmed in laboratory-reared female meadow voles (421).
These apparently contradictory results might be explained by a complex regulatory mechanism. In fact, a single administration of estradiol initially enhances (within 4 h) and subsequently suppresses (within 48 h) cell proliferation in the DG of ovariectomized female rats or meadow voles (420, 421). The increase in cell proliferation is mediated by serotonin (29), whereas the decrease is prevented by adx (422), suggesting an involvement of corticosterone. However, although corticosterone-induced regulation of cell birth certainly involves NMDA receptors, estradiol influences on cell proliferation are not mediated by these receptors (420). Finally, estrogen may act directly on estrogen receptors subtype 
(ER ) present on hippocampal precursors (435) or through IGF-I receptors (see sect. IID2C).
